Reconstruction of North American Drainage Basins and River Discharge Since the Last Glacial Maximum

Home , Barnes Ice Cap, Glacial Lake Columbia, Kankakee Torrent, Lake Agassiz, Lake Missoula, Lake Ojibway

Earth Surf. Dynam., 4, 831–869, 2016 www.earth-surf-dynam.net/4/831/2016/ doi:10.5194/esurf-4-831-2016 © Author(s) 2016. CC Attribution 3.0 License.

Reconstruction of North American drainage basins and river discharge since the Last Glacial Maximum

Andrew D. Wickert Department of Earth Sciences and Saint Anthony Falls Laboratory, University of Minnesota, Minneapolis, MN, USA Correspondence to: Andrew D. Wickert ([email protected])

Received: 1 February 2016 – Published in Earth Surf. Dynam. Discuss.: 17 February 2016 Revised: 6 October 2016 – Accepted: 12 October 2016 – Published: 8 November 2016

Abstract. Over the last glacial cycle, ice sheets and the resultant glacial isostatic adjustment (GIA) rearranged river systems. As these riverine threads that tied the ice sheets to the sea were stretched, severed, and restructured, they also shrank and swelled with the pulse of meltwater inputs and time-varying drainage basin areas, and some- times delivered enough meltwater to the oceans in the right places to influence global climate. Here I present a general method to compute past river flow paths, drainage basin geometries, and river discharges, by combining models of past ice sheets, glacial isostatic adjustment, and climate. The result is a time series of synthetic paleohydrographs and drainage basin maps from the Last Glacial Maximum to present for nine major drainage basins – the Mississippi, Rio Grande, Colorado, Columbia, Mackenzie, Hudson Bay, Saint Lawrence, Hudson, and Susquehanna/Chesapeake Bay. These are based on five published reconstructions of the North American ice sheets. I compare these maps with drainage reconstructions and discharge histories based on a review of obser- vational evidence, including river deposits and terraces, isotopic records, mineral provenance markers, glacial moraine histories, and evidence of ice stream and tunnel valley flow directions. The sharp boundaries of the reconstructed past drainage basins complement the flexurally smoothed GIA signal that is more often used to validate ice-sheet reconstructions, and provide a complementary framework to reduce nonuniqueness in model reconstructions of the North American ice-sheet complex.

1 Introduction therefore patterns of drainage-basin-scale water balance and river discharge to the oceans. At the time of Last Glacial Maximum (LGM) ice extent, Reconstructions of past ice-sheet thickness have prolif- ca. 30 to 19.5 ka (Clark et al., 2009; Lambeck et al., 2014), erated (e.g., Tushingham and Peltier, 1991; Marshall and North American drainage configurations and river discharge Clarke, 1999; Lambeck et al., 2002; Peltier, 2004; Tarasov were dramatically different from those at present. Ice ad- and Peltier, 2006; Tarasov et al., 2012; Gregoire et al., vance rerouted rivers (e.g., Ridge, 1997; Curry, 1998; Blu- 2012; Argus et al., 2014; Peltier et al., 2015), but are eval- mentritt et al., 2009), and ice streams fed temporarily en- uated using limited (Tarasov and Peltier, 2006; Tarasov larged drainage basins (Dyke and Prest, 1987; Patterson, et al., 2012) to no (all others) geologic evidence for past 1998; Margold et al., 2014, 2015). The growth of the drainage patterns. Most of these (all but Marshall and Clarke, ice sheets changed atmospheric circulation (Kutzbach and 1999, and Gregoire et al., 2012) were tuned or calibrated Wright, 1985; COHMAP members, 1988; Bromwich et al., to terminal moraine positions and records of glacial iso- 2004; Kim et al., 2008; Ullman et al., 2014) by generat- static adjustment, with the latter being a response with a ing up to 3.5–4 km of high-albedo ice-surface topography characteristic two-dimensional flexural half-wavelength of (e.g., Peltier, 2004; Tarasov et al., 2012; Peltier et al., 2015; 500–850 km over the high-flexural-rigidity Canadian Shield Ivanovic´ et al., 2016a) that influenced global climate and provence that was covered by the ice sheet (Vening Meinesz,

Published by Copernicus Publications on behalf of the European Geosciences Union. 832 A. D. Wickert: North American deglacial water routing

1931; Kirby and Swain, 2009; Tesauro et al., 2012). G12 resentations of the modern drainage network wholly or par- (Gregoire et al., 2012), which is discussed in detail here, was tially in place of glacial-stage drainage basins (e.g., Blum built to roughly match ice-sheet outlines and evolve more et al., 2000; Sionneau et al., 2008; Montero-Serrano et al., freely. Additional geological data can help to constrain past 2009; Sionneau et al., 2010; Kujau et al., 2010; Lewis and ice-sheet geometries (see Stokes et al., 2015, who review and Teller, 2006; Tripsanas et al., 2007, 2014; Rittenour et al., motivate data–model integration efforts), but until now, sur- 2003, 2005, 2007; Knox, 2007) that help to propagate a face water flow paths have not been used to critically eval- lack of consciousness about the continental-scale hydrologic uate the suite of proposed past ice sheets. Flow-routing cal- changes that occurred, even in cases when the study acknowl- culations (e.g., Metz et al., 2011; Schwanghart and Scher- edges in the text that past drainage pathways were differ- ler, 2014) are deterministic and can be used to produce ent. The best current Cenozoic history of Mississippi River time series of drainage pathways predicted by each recon- drainage does not attempt to place its northern drainage basin structed ice sheet and its associated glacial isostatic adjust- boundary during the Pleistocene (Galloway et al., 2011). ment (GIA) history, forming the necessary links between Model-based studies likewise either test a range of possible each reconstructed ice sheet and measurable geomorphic drainage basin areas (Overeem et al., 2005) or assume that and sedimentary records. Furthermore, the discrete bound- the rivers have remained unchanged (Roberts et al., 2012). aries of drainage basins are sensitive to ice geometry in a In the former case, this can lead to less-precise calcu- way that is distinct from the broad and more diffuse GIA lations of sediment transport and deposition. In the lat- response, and therefore can provide a partly independent ter case, this violates the underlying assumptions used to check on the accuracy of any given ice-sheet reconstruction compute erosion and infer spatially distributed uplift. Pa- (Wickert et al., 2013). leoclimate general circulation model (GCM) reconstruc- On a global scale, climate change during deglacia- tions (e.g., Liu et al., 2009; He et al., 2013) maintain static tion may have been forced by meltwater delivery to drainage basins in spite of climate sensitivity to patterns of the oceans, especially at the time of the Younger Dryas meltwater routing (Condron and Winsor, 2012), and this has (e.g., Broecker et al., 1989). The global climate system is just begun to be addressed (Ivanovic´ et al., 2014, 2016a,b). highly sensitive to the locations of meltwater inputs to the All of these scenarios highlight the need for an improved oceans (e.g., Tarasov and Peltier, 2005; Murton et al., 2010; community-wide understanding of the ice-age effects on the Condron and Winsor, 2012; Carlson and Clark, 2012). Ge- continental-scale river systems of North America that pro- ologic data provide a timeline of meltwater delivery to one vides the necessary backdrop to understand the present-day basin or another (e.g., Ridge, 1997; Licciardi et al., 1999; rivers, their origins, and their evolution. Clark et al., 2001; Flower et al., 2004; Knox, 2007; Rittenour Here, I focus primarily on nine North American river et al., 2007; Carlson et al., 2009; Breckenridge and Johnson, basins. These are (counterclockwise from north) the Macken- 2009; Murton et al., 2010; Hoffman et al., 2012; Williams zie River, Columbia River, Colorado River, Rio Grande, Mis- et al., 2012; Breckenridge, 2015), with some types of data, sissippi River, Susquehanna River/Chesapeake Bay, Hud- such as oxygen isotopes, being able to be analyzed to pro- son River, Saint Lawrence River, and Hudson Bay/Hudson duce reconstructions of meltwater discharge (Moore et al., Strait. These catchments include those that were dramatically 2000; Carlson et al., 2007a; Carlson, 2009; Obbink et al., and directly impacted by Quaternary glaciation and those 2010; Wickert et al., 2013). The relationships between geo- that were far from the ice sheets, and together they consti- logical data in different parts of the continent must be con- tute a reasonably representative set of rivers from which the nected by conservation of ice-sheet mass and glaciological continental-scale hydrologic impact of glaciation on North continuity of ice flow. Continental-scale meltwater routing American drainage basins can be surmised. calculations that connect data and models address both of The present work is inspired by the dual and con- these concerns by using quantified ice-sheet thickness and nected needs for (1) an appropriate mapping tool for ensuring that water that is not sent to one basin is sent to data–model intercomparison and (2) maps of the past another. drainage basins of North America to inform geologic In general, there exists a lack of recognition of the im- studies. I first present a standardized and automated portance of Pleistocene drainage rearrangement on river sys- method to extract past drainage pathways and river dis- tems. In spite of the broad knowledge that river systems and charges from combinations of ice-sheet, GIA (Mitrovica drainage basins have changed significantly since the LGM and Milne, 2003; Kendall et al., 2005), and climate (Upham, 1883; Bell, 1889; Clayton and Moran, 1982; Dyke (Liu et al., 2009; He, 2011) reconstructions. I then apply this and Prest, 1987; Wright, 1987; Teller, 1990a,b; Marshall and method to five ice-sheet models: the old but still-relevant Clarke, 1999; Patterson, 1997, 1998; Licciardi et al., 1999; ICE-3G (Tushingham and Peltier, 1991); the “industry stan- Overeem et al., 2005; Tarasov and Peltier, 2006; Wickert, dard” ICE-5G (Peltier, 2004); the ANU model, developed 2014; Ullman et al., 2014; Margold et al., 2014; Ullman et al., by a separate research group (Lambeck et al., 2002); the ice- 2015; Margold et al., 2015), many studies do not include any physics-based “G12” model (Gregoire et al., 2012); and the clear picture of drainage basin evolution. This results in rep- new ICE-6G (Argus et al., 2014; Peltier et al., 2015). I

Earth Surf. Dynam., 4, 831–869, 2016 www.earth-surf-dynam.net/4/831/2016/ A. D. Wickert: North American deglacial water routing 833 then compare the resultant drainage basin evolution and synthetic paleohydrographs against onshore and offshore geologic data, including data-driven reconstructions of North American drainage basins.

2 Methods: automated drainage reconstruction

River systems in North America have responded to synchronous and linked changes in climate, ice extent, and sea level, with the latter including GIA-induced variability. Ad- dressing this connected climate–ice–solid-Earth system is necessary to accurately reconstruct drainage. Ice advances Figure 1. Meltwater and meteoric (precipitation minus evapotran- and retreats blocked drainage pathways and rerouted rivers spiration) water are routed to the coast, across a surface formed by (Tight, 1903; Ver Steeg, 1946; Bluemle, 1972; Teller, 1973; the sum of present-day topography, glacial isostatic adjustment, and Licciardi et al., 1999; Teller et al., 2002; Stanford, 2010; ice-sheet thickness. Carlson and Clark, 2012; Prince and Spotila, 2013). Ice- sheet retreat left behind large isostatic depressions that held proglacial lakes, such as the massive glacial Lake Agassiz the Python source code to automate these analyses has been (e.g., Breckenridge, 2015), and strongly affected drainage released under the GNU General Public License (GPL) ver- patterns (e.g., Fisher and Souch, 1998). Even the subtle sion 3 and made available to the public (Sect.7). ( ∼ 20–60 m) forebulges of continental ice sheets may have played major roles in drainage rearrangement (Anderson, 2.1 Surface elevations for flow routing 1988; Stanford et al., 2016). This is especially true of the gen- Water and ice flow downslope following a routing surface, erally low-relief North American cratonal interior, where ice Z , that is given by sheets and GIA have been significant drivers of large-scale r topographic slopes. Zr = Zm,b + Hi + 1Zb. (1) The drainage basin and paleohydrologic reconstructions are a combination of four inputs: (1) a digital elevation model Zm,b is the reference (i.e., modern) land-surface elevation (DEM) with high enough resolution to resolve the valleys of (subscript “m,b” stands for “modern glacier bed” or “mod- major river systems, (2) a prescribed ice-sheet history, (3) a ern bare (i.e., ice-free) ground”), Hi is the ice thickness, and model of the GIA response to that ice history, and (4) a past 1Zb represents changes between the past and modern land- water balance based on GCM outputs that have been offset surface (glacier-bed) elevation. Together, Zm,b + 1Zb give to fit modern data at the present time step (Fig.1). These fac- the past glacier-bed elevation field. When this (Zm,b + 1Zb) tors together allow us to generate synthetic river discharge surface is summed with ice thickness, Hi, the resulting rout- histories that can be compared with geologic data. ing surface, Zr, combines ice-free land-surface elevations This analysis produced three major end products. The first and the elevations of the surface of the ice sheet. The ice- is a gridded, spatially distributed map of estimated river dis- sheet surface is chosen to drive the flow-routing calculations charges to the sea since the LGM, expressed as centennial because this is what generates the driving stress for ice flow to millennial means depending on the resolution of the in- (see Cuffey and Paterson, 2010, Sect. 8.2.1) and therefore put ice-sheet model. The second is a coarsened version of sets ice divides and meltwater pathways. If one wanted to the first, designed for coupling with GCMs to investigate the compute subglacial water routing instead, an alternate ap- feedbacks between ice melt, ocean circulation, and climate proach would be to drive flow by the subglacial pressure (see Ivanovic´ et al., 2014, 2016b). The third is a set of major gradient to simulate meltwater routing through subglacial drainage basin extents and a time series of their discharges streams and tunnel channels (e.g., Wright, 1973), and this into the ocean. Here, I focus on this third product, as con- could be accomplished by multiplying Hi by ρi/ρw, where sidering these specific drainage basins simplifies an effect the latter are the densities of ice and water, respectively (e.g., spread across tens of thousands of pixels into nine discrete Livingstone et al., 2015). entities for which there exists a sufficient geologic record to compare model outputs and data. 2.1.1 DEM and grid resolution The drainage analyses were performed using GRASS GIS (Neteler et al., 2012; GRASS Development Team, 2015), a Zb is modern land-surface elevation. The primary modern el- free and open-source GIS software that has efficient data evation data set for flow routing is the GEBCO_08 30 arcsec management, computational, and hydrologic tools. Any se- DEM (British Oceanographic Data Centre, 2010). GEBCO ries of operations within GRASS GIS may be scripted, and provides the highest-resolution combined topography and www.earth-surf-dynam.net/4/831/2016/ Earth Surf. Dynam., 4, 831–869, 2016 834 A. D. Wickert: North American deglacial water routing bathymetry data sets currently available. Thirty arc seconds plateaus and several-hundred-meter cliffs at their termini that corresponds to a north–south cell size of 927 m and an east– would otherwise introduce artifacts in the flow-routing calcu- west cell size that ranges from 895 m at 15◦ N, the southern lations. For a 1◦ grid, this requires five iterations and leads to extent of the analysis region, to 81 m at 85◦ N, the northern smoothing over a half-degree distance (∼ 40–50 km) outside extent. The GEBCO_08 grid is used in all areas except be- of a given cell. While this somewhat smears the ice-sheet neath the Greenland Ice Sheet, where the subglacial topog- boundary (see, e.g., “ICE-6G” vs. “From data” in Figs.7 raphy is interpolated from the etopo1 1 arcmin global topo- and8), it does not significantly change its fit to the published graphic data set (Amante and Eakins, 2009). ice-margin chronology (Dyke et al., 2003; Dyke, 2004). This sub-kilometer resolution over the ice-free Earth is re- I calculate deglacial drainage using five ice models, four quired to resolve flow inside the kilometer-scale valleys of of which were created based on geophysical data, and one of the major rivers that drain North America, and is therefore which was generated using a numerical model of ice-sheet critical to accurately calculate river networks. More coarsely physics. In the former, ice-sheet thickness evolution was de- gridded drainage routing can send water from the modern termined by inverting for global sea-level data by combining upper Missouri River basin, for example, towards the Arctic (1) the total water mass contained within the ice sheets, caus- Ocean. ing global mean sea-level fall, and (2) calculations of the GIA However, even these high-resolution calculations exhibit process, which produces deviations from mean sea-level his- some small deviations from mapped drainage basin extents. tory. Most notable among these deviations is the local defor- Examples of this, such as in the Arrowhead region of Min- mation of the lithosphere and mantle due to the weight of the nesota, where the Saint Louis River is believed to have ice, which leads to postglacial rebound following ice retreat. progressively captured upper Mississippi drainage near Sa- Ongoing postglacial rebound is a key indicator of the pres- vanna Portage (van Hise and Leith, 1911), indicate that still- ence of major past ice sheets, but postglacial rebound rates higher resolution elevation data will continue to improve are a function of both time-evolving past ice-sheet thickness flow-routing calculations near headwater streams, a scale de- and solid-Earth (lithosphere and mantle) rheology. That this pendence given by the relationship between channel geom- one rate is a function of two only partly constrained vari- etry, river discharge, and drainage area (Leopold and Mad- ables introduces an element of uncertainty into GIA-based dock, 1953). inversions (Mitrovica, 1996) such as the ICE-nG and ANU Raster grid cell area, Ac, is a function of latitude, and this models. Ice physics, on the other hand, are most sensitive to is an important conversion factor for changing scalar quan- the paleoclimate forcing and basal conditions (e.g., topog- tities in cells, such as ice thickness, into a measure of vol- raphy, roughness, sedimentary cover, geothermal heat flux) ume. Using the simplifying assumption that Earth is perfectly (Tarasov and Peltier, 2006; Cuffey and Paterson, 2010). The spherical, four geophysically based ice-sheet reconstructions investi- gated here, listed with their associated solid-Earth rheology π π Ac = R⊕ ◦ δθ R⊕ sin(θ) ◦ δϕ . (2) models, are ICE-3G/VM1 (Tushingham and Peltier, 1991), 180 180 ANU (Lambeck et al., 2002), ICE-5G/VM2 (Peltier, 2004), Here, R⊕ is the mean radius of Earth, θ is colatitude (in de- and ICE-6G_C/VM5a (Argus et al., 2014; Peltier et al., grees), and ϕ is east longitude (i.e., degrees east of Green- 2015). The one ice-physics-based reconstruction considered wich). δθ is cell north–south extent and δϕ is cell east–west is that of (Gregoire et al., 2012), which was allowed to evolve extent; in the simulations presented here, both of these ex- more freely than the tuned ice-physics-based model recon- tents equal 30 arcsec. These areas are computed at a latitude, structions of Tarasov and Peltier(2006) and Tarasov et al. λ, corresponding to the north–south midpoint of the cell, with (2012). the implicit assumption that the cells are small enough that ICE-3G (Tushingham and Peltier, 1991) was one of the north–south difference in cell size can be approximated the earliest widely used ice-sheet reconstructions. It was to be linear. For the GEBCO_08 30 arcsec grid that is used constructed to constrain ice thickness as much as possi- here, this approximation generates errors of < 0.3 %. ble with measurements of glacial isostatic adjustment from radiocarbon-dated sea-level markers. While it does not include enough ice to reproduce the 2.1.2 Ice-sheet history full magnitude of the observed LGM global sea-level fall Hi, ice-sheet thickness, is provided by pre-generated ice- (Austermann et al., 2013; Lambeck et al., 2014), it provides sheet reconstructions. These are produced at a coarser res- a counterpoint to the more recent ice-sheet reconstructions, olution than the 30 arcsec flow-routing grid and therefore are all of which include a much more massive North American interpolated using an iterative nearest-neighbor approach to ice-sheet complex (Fig.2). remove stepwise discontinuities. This smoothing works suc- ICE-5G (Peltier, 2004) has been often considered the “in- cessively from one-third the native resolution of the ice-sheet dustry standard” global ice-sheet and GIA reconstruction. It reconstruction to 30 arcsec, and is necessary because some was also built by combining geological data and geophysical coarse ice-sheet reconstructions contain multi-cell interior inversions, and is an update to ICE-3G. It contains a very

Earth Surf. Dynam., 4, 831–869, 2016 www.earth-surf-dynam.net/4/831/2016/ A. D. Wickert: North American deglacial water routing 835

implausible as the ice sheet would slide and thin over the soft Western Interior Seaway sediments east of the Cana- dian Rockies. Nevertheless, ICE-5G remains widely used and matches many observations of GIA and sea level (Peltier, 2004). ICE-6G (here, shorthand for ICE-6G_C) (Argus et al., 2014; Peltier et al., 2015) is the successor to ICE-5G. It has been shown to improve agreement with present-day GIA measurements, and was built to match geologic evidence for (a) the locations of the ice domes and ice-sheet outlines (Patter- son, 1998; Dyke et al., 2003; Dyke, 2004). As ICE-6G is so new, it has not been extensively tested outside of the group that developed it (Argus and Peltier, (b) 2010; Argus et al., 2014; Peltier et al., 2015). Lambeck et al.(2002) developed the Australian National University (ANU) ice-sheet model, also based on field data constraining sea level and GIA. This model differs from ICE- 3G and ICE-5G in that it was developed regionally, and these regions were later combined into a more global picture of ice-sheet evolution. To produce the G12 ice model, Gregoire et al.(2012) sim- ulated North American ice-sheet evolution from the LGM to present using the Glimmer Community Ice-Sheet Model (Glimmer–CISM) (Rutt et al., 2009). Glimmer–CISM is a Figure 2. Ice volume and meltwater discharge histories over the thermomechanical model of ice deformation and dynam- North American computational region for each of the examined ice- ics that uses the shallow ice approximation for its driv- sheet reconstructions (colors), along with changes in global ice vol- ing stresses. While the shallow ice approximation neglects ume from Lambeck et al.(2014) (gray). (a) Cumulative change in higher-order (i.e., less important) stresses, it generally is an ice volume over time. Volumes for the five ice-sheet reconstruc- acceptable approximation to solutions that do include higher- tions evaluated here are summed over the computational domain, order terms, and its computational efficiency makes it a vi- which contains the entire North American ice-sheet complex and able option for long-term simulations (Rutt et al., 2009), such part of Greenland. The Lambeck et al.(2014) ice-volume curve as the Gregoire et al.(2012) LGM–present study. The G12 is global but has been shifted to match ICE-6G at the 0 ka (i.e., reconstruction features the collapse of the ice saddle between modern) time step for easier comparison with the deglacial North American curves. These ice volumes are presented in meters sea- the Laurentide and Cordilleran ice sheets around 11.6 ka. Ge- level equivalent (SLE), assuming a constant ocean surface area of ological data, on the other hand, place the major phase of 3.62 × 108 km2 (after Marshall and Clarke, 1999). (b) Total melt- saddle collapse during the Bølling–Allerød warm period, ca. water discharge over the computational domain for each ice-sheet 14.5–12.9 ka (Dyke et al., 2003; Dyke, 2004; Williams et al., reconstruction. ICE-5G and ICE-6G have a noticeable peak during 2012), and possibly beginning as early as ∼ 17.0 ka (Catto MWP-1A, 14.65–14.31 ka (Deschamps et al., 2012). Though the et al., 1996; Clague and James, 2002; Carlson and Clark, timescale of their numerical model is shifted, Gregoire et al.(2012) 2012). While Gregoire et al.(2012) shifted their chronol- also consider their meltwater peak, due to collapse of the ice sad- ogy such that peak saddle collapse occurred ∼ 14.5 ka, dur- dle between the Cordilleran and Laurentide ice sheets, to represent ing Meltwater Pulse 1A (MWP-1A) (Fairbanks, 1989; De- MWP-1A. The global-scale Lambeck et al.(2014) sea-level curve schamps et al., 2012), I use their model ages as direct geo- includes a meltwater peak ∼ 16.5–15 ka, which encompasses Hein- logic age, such that the results presented here are linked in rich Event 1B (15.9–15.7 ka) (Gil et al., 2015) and significant ice- sheet melt (12.4–6.7 ka), during the first half of the Holocene. The time with the effects of their climate forcings on their ice vertical axis is in sverdrups, where 1 Sv is 106 m3 s−1. sheet. Most importantly, the ice dynamics contained within Glimmer–CISM provide a set of tools to match reality that place G12 in a different category than the GIA-based models. thick north–south-oriented ice dome over western Canada (Argus and Peltier, 2010) that is not located where glacial 2.1.3 Change in land-surface elevation: glacial isostatic geological evidence would place the Keewatin dome. This adjustment causes its meltwater discharge to the coast to fail to match isotopic records from the Mississippi River drainage basin 1Zb, change in land-surface (glacier-bed) elevation, is here (Wickert et al., 2013), which, in any case, is glaciologically provided by the combination of factors that change local rel- www.earth-surf-dynam.net/4/831/2016/ Earth Surf. Dynam., 4, 831–869, 2016 836 A. D. Wickert: North American deglacial water routing ative sea level. At the LGM, ice-volume equivalent global with a viscosity of 1022 Pa s. The ANU solid-Earth model has mean sea level was 130–135 m lower than present, but local a higher viscosity contrast between the upper and lower man- relative sea level was highly variable due to the effects of tle than VM2, placing it in better agreement with geophys- GIA (Austermann et al., 2013; Lambeck et al., 2014). GIA ical inferences of the upper mantle–lower mantle viscosity comprises deformational, gravitational, and rotational effects contrast based on observations of the long-wavelength geoid of ice-age loading. The deformational component is the lo- (Hager and Richards, 1989), and employs a 65 km thick elas- cal flexural isostatic response to changes in ice and water tic lithosphere. One may note that all of the lithospheric elas- loading. The gravitational component is in response to redis- tic thicknesses are characteristic of cratonal regions; this is tribution of mass (and hence gravitational potential) across because the Laurentide and Fennoscandian ice sheets nucle- the surface of Earth, and this is amplified by the water self- ated and grew across old, thick cratonal lithosphere, and a gravitation feedback (e.g., Mitrovica and Milne, 2003). The cratonal value for elastic thickness therefore provides the rotational component results from perturbations to Earth’s best fit to near-field data that record GIA, whose local de- moment of inertia due to global mass redistribution, and pro- formational component is high-amplitude relative to the ro- duces true polar wander that causes the equatorial bulge to tational and gravitational components. migrate, first in the form of seawater and later in the form of I initiate the GIA modeling of G12 in equilibrium at the solid-Earth material. These three components feed back into LGM, and ANU slightly before the LGM. This is a poten- one another and therefore are solved numerically through it- tial source of error: while the sea-level records of Lambeck eration. et al.(2014) indicate that global ice mass grew only slightly The gravitationally self-consistent sea-level theory applied between 30 and 20 ka, providing enough time to approach here and its numerical implementation are described by isostatic equilibrium, this is only ∼ 2 e-folding timescales to- Mitrovica and Milne(2003) and Kendall et al.(2005), re- wards isostatic equilibrium based on data from the Ångerman spectively, who calculate the GIA effects of changing ice River and Hudson Bay (Mitrovica and Forte, 2004; Nord- and water masses. The calculations are performed using a man et al., 2015). Therefore, it is very likely that the North pseudo-spectral algorithm that, for this work, is truncated at American ice-sheet complex was not yet in isostatic equi- spherical harmonic degree and order 256. This satisfies the librium at the LGM. ICE-5G includes less well-constrained Nyquist flexural wavenumber for the solid-Earth models em- ice-sheet growth since 120 ka. The starting point for ICE-6G ployed, meaning that the wavelength of the spectral repre- is during the last interglacial (MIS 5e), when global mean sentation is always less than half of the flexural wavelength sea level peaked 5.5–7.5 m above its present level (Hay et al., (Vening Meinesz, 1931) of the lithosphere in the solid-Earth 2014) though with some significant regional variability, as model. Satisfaction of this condition allows the solutions to indicated by ancient beaches and modeling studies (Hearty be interpolated to an arbitrarily high resolution without con- et al., 2007; Rohling et al., 2007; Dutton and Lambeck, 2012; cerns about aliasing. Hay et al., 2014; Dutton et al., 2015). ICE-3G was initiated at ICE-3G, ICE-5G, ICE-6G, and ANU were each calibrated the last interglacial as well, with ice being grown in the same to match sea-level data based on specific 1-D models of pattern as it retreated, except more slowly and in reverse. solid-Earth viscoelasticity (Tushingham and Peltier, 1991; Our flow-routing calculations neglect geomorphic change Lambeck et al., 2002; Peltier, 2004; Argus et al., 2014; for three major reasons. First, rerouting of Lake Agassiz Peltier et al., 2015); Argus and Peltier(2010, p. 699) provide outflow occurred as a precursor to spillway incision (Teller a succinct review of the VMn models. I use each ice recon- et al., 2002), meaning that ice-sheet geometry and GIA struction’s respective spherically symmetric Earth model to are the primary drivers of drainage reorganization. Second, simulate its GIA response to these loads – VM1 for ICE-3G, while process-based landscape evolution models exist (see VM2 for ICE-5G, VM5a for ICE-6G, and an unnamed model Tucker and Whipple, 2002; Willgoose, 2005; Tucker and for ANU. G12 was built using a simplified model of iso- Hancock, 2010), the thresholded and highly nonlinear re- static response to glacial loading, and was run only for North lationships between flow, erosion, sediment transport, and America. I pair it with VM2 and a posteriori enforce a total deposition, means that these models may predict a very global ice-sheet volume (and therefore ice-volume equiva- wide range of potential landscape responses with sufficient lent global mean sea-level history) equal to that of ICE-5G. uncertainty as to reduce their predictive capability. This Among these solid-Earth models, VM1 is a two-layer model would leave changing the DEMs by hand as an option with a viscosity that increases at the transition zone. VM2 (as was done by Tarasov and Peltier, 2005), but considering includes many layers, with a large increase in viscosity be- incomplete knowledge of past topography and my desire to, low the 660 km discontinuity and gradually increasing vis- at least at this stage, separate the calculations from the field cosity with depth thereafter in the lower mantle; these layers work against which they are checked, I choose to use modern underly a 90 km thick elastic lithosphere in the most recent topography while acknowledging that this is an approxima- version of VM2, which is used here (Peltier, 2004). Below tion. 100 km, VM5a is a simplification of VM2; above 100 km, it includes a 60 km thick lithosphere over a 40 km thick layer

Earth Surf. Dynam., 4, 831–869, 2016 www.earth-surf-dynam.net/4/831/2016/ A. D. Wickert: North American deglacial water routing 837

2.2 Water balance: precipitation, evapotranspiration,

P ET (21 ka) P ET (0 ka) 1 and ice melt − − − 604 − Each cell in the ﬂow-routing grid domain can act as a source 483 of water to continental drainage systems. Water balance over 362 each cell is a sum of precipitation gains, P ; evapotranspiration losses, ET; and changes in water-equivalent thickness of 242 the reconstructed ice sheet, (ρ /ρ )(1H /1t). When multi- i w i 121 plied by the cell area, Ac (Eq.2), the result is an expression for volumetric incoming water discharge, Qin (Eq.3), that is 0 then routed across the continent. -96 ρi 1Hi Qin = Ac P − ET − (3) -192 ρw 1t -287

2.2.1 Global precipitation and evapotranspiration -383

Precipitation and evapotranspiration inputs changed through -479 time (Fig.3), and precipitation minus evapotranspiration yr[mm difference evapotransp. minus Precip. ] (P − ET) provides the amount of meteoric water that can be Figure 3. Modeled changes precipitation minus evapotranspiration routed through river systems to the coast. Changes in P −ET between the LGM and present. Modeled changes precipitation mi- through time were calculated by Liu et al.(2009) and He nus evapotranspiration between the LGM and present. These dif- (2011) using a continuous LGM (22 ka) to present run of the ferences were calculated with the TraCE-21K continuous LGM to National Center for Atmospheric Research (NCAR) Com- present paleoclimate run (Liu et al., 2009; He, 2011) of the CCSM3 munity Climate System Model version 3 (CCSM3) (Collins GCM (Collins et al., 2006). They are combined with ice mass bal- et al., 2006). This simulation, called TraCE-21K, was run at ance to produce water inputs for computed paleodischarges. 3.5◦ resolution, using ICE-5G/VM2 for its ice-sheet and pa- leotopography input. Therefore, the only fully self-consistent uct provided precipitation over the global continents except calculations presented here are from the ICE-5G/VM2 re- Antarctica (expanded from the work of Harris et al., 2014), construction. I spherically interpolated (following Dierckx, and this was smoothed and spherically interpolated (fol- 1993) the results to 0.25◦ (15 arcmin) resolution to match the lowing Dierckx, 1993) to 0.25◦ resolution. The HOAPS-3 assembled modern data products (below). satellite-derived precipitation data product (0.25◦ resolution) The low-resolution CCSM3 contains systematic biases was used for the oceans north of the Antarctic region (An- that produce a ∼ 1 mm day−1 positive precipitation anomaly dersson et al., 2010). Gaps of 2◦ or less between these data in western North America and a ∼ 1 mm day−1 negative pre- sets were filled by an interpolation of them. The remainder cipitation anomaly in eastern North America (Yeager et al., of the globe, primarily the Antarctic region, was filled by the 2006). To correct for this and other possible model biases, CMAP reanalysis data set (Xie and Arkin, 1997), interpo- I assembled modern precipitation and evapotranspiration lated from 2.5◦ resolution to 0.25◦ resolution. fields, computed the difference between each time step of Our modern evapotranspiration rates for the continents TraCE-21K and its modern time step, and summed the mod- (except Greenland and Antarctica) are based on calculations ern compilations with the computed TraCE-21K difference using the MODerate-resolution Imaging Spectroradiometer from modern. I could find no combined land–ocean evap- (MODIS) instruments that are mounted on the Aqua and otranspiration product, and combined land–ocean precipita- Terra satellites (Mu et al., 2007, 2011), available at 30 arcsec tion products were coarse, often at 2.5◦ resolution (∼ 280 km resolution. The oceanic evaporation data are from HOAPS-3 cells at the equator) (Xie and Arkin, 1997; Chen et al., and at 0.25◦ resolution (Andersson et al., 2010). Analysis of 2002; Adler et al., 2003). Therefore, I generated my own the HOAPS-3 data product shows that it has an ∼ 10 % error, global data product compilations as mean rates from 2000 to with a global 0.2–0.5 mm day−1 excess evaporation (Ander- 2004, for consistency with prior work (Wickert et al., 2013), sson et al., 2011). These data sets were interpolated over the at 0.25◦ resolution. While this study centers on North Amer- ice-free land surface and global oceans. Evapotranspiration ica, this global climate data product can be used to extend rates for Greenland and Antarctica were taken from the 0 ka these methods to other regions or the full globe (see Ivanovic´ time step of the TraCE-21K simulation (Liu et al., 2009; He, et al., 2014, 2016b). These compiled global grids of mean 2011), spherically interpolated to 0.25◦ resolution. Any neg- precipitation and evapotranspiration are publicly available ative evapotranspiration rates generated by the interpolation (see Sect.7). were set to 0. I derived the precipitation data by stitching together three data products. The CRU TS3.21 0.5◦ precipitation prod- www.earth-surf-dynam.net/4/831/2016/ Earth Surf. Dynam., 4, 831–869, 2016 838 A. D. Wickert: North American deglacial water routing

2.2.2 Ice melt and central differencing here (Tushingham and Peltier, 1991; Lambeck et al., 2002; Peltier, 2004; Gregoire et al., 2012; Peltier et al., 2015) do Ice-sheet inputs to runoff are determined by differencing ice- not. This means that these models implicitly include the lake- sheet thicknesses at adjacent time steps for each of the ice water volume within the ice mass. This deficiency affects the models. These differences in ice-sheet thickness are con- ability of the models to match the observed record of ice- verted to water discharges by (1) multiplying them by the marginal geometry, as ice must be used in the models to rep- DEM cell size (Eq.2), (2) multiplying them by 0.917 as a resent the surface loads of these large lakes in the geologic conversion factor between ice and water density, and (3) di- past, and removes the potential effects of the lakes in driving viding them by the time elapsed (Eq.3). ice-sheet retreat in mechanistic models. In the present work, Flow routing and drainage basins can be calculated only water is simply routed across proglacial-lake-producing de- on time steps when the ice-sheet and GIA models provide pressions using a least-cost path algorithm (see Sect. 2.3). surface topography (Sect. 2.3, below), but the ice-sheet contribution to runoff requires ice-sheet thickness to be dif- 2.3 Continental flow routing and accumulation ferenced. This requires a choice of whether to allow central differencing and numerical diffusion in space (i.e., flow After computing the amount of flow each cell contributes routing) or time (i.e., melt rate). Melt rates change more to runoff, I used the time-evolving grids of surface eleva- gradually than near-instantaneous drainage rerouting events, tion to compute drainage patterns across North America. At and these rerouting events are critical to understanding cli- each time step, I first defined the ocean margin that sets the mate impacts of ice melt (e.g., Broecker et al., 1989; Clark downstream boundary for terrestrial water flow. I then routed et al., 2001; Tarasov and Peltier, 2006; Condron and Winsor, flow down the Zr (Eq.1) surface while computing accu- 2012). Therefore, I prefer numerical diffusion in time, and mulated discharge as (1) meteoric inputs only, (2) ice-sheet allow the gridded melt rate at each flow-routing time step to inputs only, and (3) combined total discharge. I computed equal the time-weighted averages of the melt rates calculated this flow routing and accumulation using “r.watershed”, the before and after it. high-efficiency least-cost path search algorithm of Metz et al. Time steps vary by model. ICE-3G (Tushingham and (2011), which is integrated into GRASS GIS (Metz et al., Peltier, 1991) and ANU (Lambeck et al., 2002) use irreg- 2011; Neteler et al., 2012; GRASS Development Team, ular time steps separated by ∼ 1000 years (ICE-3G) and 2015). ∼ 500 years (ANU). Time steps for ICE-5G are 1000 years from 32 to 17 ka, and 500 years thereafter (Peltier, 2004). All 2.3.1 Defining continent, ocean, and shore time steps for ICE-6G (Argus et al., 2014; Peltier et al., 2015) are 500 years. Model outputs from Gregoire et al.(2012) Shorelines and topographic elevations changed dramatically were provided at 10-year intervals, with GIA computed at since the LGM, thus requiring redefinition of the coastline on uniform 100-year time steps (Wickert et al., 2013), which are which river mouths appear at each time step. This could not the same as those used here for the flow routing calculations. be done by simply picking regions below sea level, because This means that the modeled discharge histories are averaged this approach would cause flow to “disappear” into regions over 100- to 1000-year time steps and blur short-term geo- such as the below-sea-level Caribou subbasin of Lake Supe- logic events that are responsible for creating major deglacial rior and the isostatically depressed moat around the retreat- spillways and deposits (e.g., Kehew and Teller, 1994; Breck- ing Laurentide Ice Sheet margin. Therefore, I developed an enridge, 2007; Murton et al., 2010). algorithm to define as “ocean” only those regions that were below sea level and contiguous with the global ocean. First, I thresholded the topography-plus-ice flow-routing map, Zr, at 2.2.3 Neglected terms: groundwater and lakes 0, to define a binary raster map of regions above and below sea level. Then, I converted this into a vector map. Using the While groundwater is a significant reservoir that affects river topological analysis built into GRASS GIS, I looked only for discharge, I assume that over the millennial timescales of areas that touched the map edge and were below sea level, interest to us, groundwater storage changes slowly enough and chose these to be ocean. The remaining area was defined that its effect on river discharge is much less important as continent and used as the flow-routing surface. than precipitation, evapotranspiration, and ice-sheet melt. Al- though ice sheets can strongly influence groundwater discharge, modeling work by Lemieux et al.(2008) shows that the maximum net deglacial groundwater discharge across the whole North American continent is only ∼ 2.5 % the modern Mississippi River discharge, and that this spike was short- lived. I do not include proglacial lakes in the continental water storage because the ice-sheet reconstructions discussed

Earth Surf. Dynam., 4, 831–869, 2016 www.earth-surf-dynam.net/4/831/2016/ A. D. Wickert: North American deglacial water routing 839

2.3.2 Flow routing and accumulation ICE-6G: 14.5 ka to 14.0 ka 1.89 1.66

The ﬂow-routing calculations produce maps of (1) drainage ] 1

1.42 − direction, (2) meteoric water discharge, (3) meltwater dis- r y charge, and (4) total water discharge. These maps are spa- 1.18 m [

0.95 tially distributed, providing 30 arcsec-resolution time series s 0.71 s e

of the centennially to millennially averaged surface water n

0.47 k c discharge that is produced by each ice-model–solid-Earth- i

0.24 h t model combination. 0.00 e c i

The ﬂow-routing algorithm I used, r.watershed (Metz

-0.58 n i et al., 2011), employs a least-cost path search algorithm. This -1.17 e g sets it apart from other ﬂow-routing tools in that it is capa- -1.75 n a ble of routing water through depressions without explicitly h -2.34 c

ﬂooding them (see, e.g., Schwanghart and Scherler, 2014). o

-2.92 e This algorithm allows for rapid calculations to be carried out -3.50 t a over a landscape that contains moving ice dams and isostatic -4.09 R depressions. -4.67 The ability of the Metz et al.(2011) algorithm to route water across local depressions also prevents both natural and Figure 4. Ice-surface mass balance. Ice-sheet mass balance is a ma- artificial (human-built or DEM artifact) depressions in the jor, and often dominant, input to calculations of North American landscape from impeding the flow routing. This is an advan- river discharge during deglaciation (Eq.3). tage because flow-routing calculations without this feature fail unless they are performed on a hydrologically corrected clare the Great Basin to be closed during part or all of the DEM (i.e., one in which the elevation is altered to produce modeled geologic past. The approach taken here is to leave the known drainage basins). Using a hydrologically corrected the flow-routing grid unmodified in order to maintain dis- DEM is a potential source of error in reconstructions of the tance between data and model (see the final paragraph in past because its actual terrain has been artificially changed Sect. 2.1.3). (e.g., Grimaldi et al., 2007) in a way that may produce unex- pected results when combined with surface deformation and ice-sheet thickness. 2.4 Distributed melt delivery to the coast In spite of these advantages, the Metz et al.(2011) method These methods also allow us to produce grids of meltwater causes internally drained regions to be amalgamated to discharge to the coast that are fully distributed in space and drainage basins that flow to the coast. In the drainage cal- time. These have been converted to GCM-input-appropriate culations performed for this work, the Great Basin is split spatial resolutions and used to connect ice melt and global between the Colorado River and Columbia River drainage ocean circulation and climate change (Ivanovic´ et al., 2014, basins. In part, this is realistic: outflow from Lake Bon- 2016b). This connection to continental runoff distinguishes neville entered Columbia River between 18.2 ± 0.3 and this distributed meltwater approach from “hosing” experi- 16.4 ± 0.2 ka, with limited and/or intermittent outflow con- ments (e.g., Meissner and Clark, 2006; Condron and Win- +0.3 ± 14 tinuing to 14.9−0.6 ka (12.6 0.15 C ka) (synthesis of God- sor, 2012), in which meltwater additions are prescribed to sey et al., 2011; McGee et al., 2012) (radiocarbon ages cali- a patch of ocean to explore the climatic impacts of a ma- brated using IntCal13 (Reimer et al., 2013) with 2σ error). In jor meltwater input event. Fully distributed GCM meltwa- part, this does not matter: a hydrologically closed basin must ter inputs that are tied to paleodrainage histories, when cou- maintain a state in which precipitation equals evapotranspira- pled with models that can appropriately handle point sources tion, and as such should not have a major effect on discharge of fresh water, create a more realistic ice-age ocean that is calculations, outside of the aforementioned period of climate better-conditioned to respond to changes in meltwater inputs change during the end of the Pleistocene. But, in part, this is and can be used to connect models of ice-sheet history with the dissatisfying result of flow-routing calculations that are their likely impacts on global climate. Code to produce these not coupled to a hydrologic model that can raise and lower GCM inputs has been released alongside this publication (see the levels of lakes in closed basins, and a fundamental prob- Sect.7). lem in any flow-routing approach that does not include lake and groundwater levels. 2.5 Drainage basins and river discharge This dilemma over the Great Basin highlights a key ques- tion regarding how much to adjust the model by hand to fit I compute time-evolving drainage patterns with a generalized data, versus how much to just allow the model to run. It algorithm that can overcome changes in shoreline geometry, would be possible to modify the flow-routing inputs to de- river mouth positions, and difficulties in producing consistent www.earth-surf-dynam.net/4/831/2016/ Earth Surf. Dynam., 4, 831–869, 2016 840 A. D. Wickert: North American deglacial water routing elevation-based flow reconstructions across the very subtle topography of the continental margins. These components are essential to connect glacial-stage rivers to their modern counterparts after 20 000 years of shoreline and topographic change due to evolving sea level and ice-sheet geometry. First, I create a map of regions where I expect the mouth(s) of each river system to be. Each “river mouth region” (Fig.5b) is defined by hand to accommodate the changing locations of river mouths with time as a function of sea level, ice cover, and GIA (Fig.6). This is especially important for drainage systems such as Hudson Strait and the Saint Lawrence River, which previously played host to ice streams and now are fully or partially inundated with seawater. The same set of river mouth regions is used for every ice-sheet and GIA reconstruction at all time steps. I then create a set of points at which each of the “major” rivers (≥ 1000 m3 s−1 reconstructed discharge) reaches the coast. Wherever one of these river mouths is located inside a named region (Fig.5), the program builds a drainage basin. Each drainage basin is labeled with the name of its respective river mouth region (e.g., “Mississippi” or “Columbia”). River mouth regions that contain more than one river mouth produce a named drainage basin that is an amalgamation of the drainage basins of all ≥ 1000 m3 s−1 rivers in the basin; this allows us, for example, to combine the Missis- sippi, Atchafalaya, and Red rivers into a single broader basin flowing towards Louisiana, and to combine all of the major modern rivers that flow into Hudson Bay into a single basin, even though their previous combined outlet in Hud- son Strait is now inundated. These river mouth regions are chosen by hand, and work only if one river mouth does not migrate into a region occupied by another river mouth at a different time. If such a dramatic change occurs, as it does for the Hudson River at 12.5 and 12.0 ka in ICE-6G and for Hudson Bay at 9.6 ka in G12 (these are the only known likely errors), it then becomes challenging to even define the river as being the same, and more reach-scale definitions may become required. If a particular river mouth region contains no Figure 5. Modern drainage systems and river mouth regions used to 3 −1 river mouths with flow ≥ 1000 m s , the program builds a reconstruct past drainage basins. (a) Modern drainage basin extents drainage basin upstream of the highest-discharge coastal cell for the rivers involved in this study, with ice and lake extents from from the flow accumulation calculation. the 1 ka time step of the maps of Dyke(2004), elevation basemap After defining the major drainage basins, I sum (1) ice- from GEBCO_08 (British Oceanographic Data Centre, 2010), and sheet melt, (2) precipitation minus evapotranspiration, and modern drainage basins from Lehner et al.(2013). (b) Modern river (3) the sum of (1) and (2) across each basin. Each sum, re- mouths alongside the river mouth regions used to define a flow path as part of a named drainage basin (Sect. 2.5). Names are also given spectively, provides ice-sheet meltwater discharge (Qi), meteoric (seasonal precipitation minus evapotranspiration) dis- in European non-English languages where they are common, alongside a non-exhaustive set of common names in local languages. charge (Q ), and total river discharge (Q = Q +Q ). Many m m i Note that the mouth of the Mississippi River lies entirely outside of these calculations provide discharges at the modern time its river mouth region. This is because the topographically routed step that are close to the modern pre-human-impact (i.e., flow, computed using the Metz et al.(2011) r.watershed algorithm, preindustrial) mean river discharge (Table1, which also in- follows the course of the Atchafalaya River; the present path of the cludes measured modern sediment discharges for reference). Mississippi near its mouth does not follow the regional topographic For some, however, the result differs somewhat due either gradient because it is held in its old course by the Old River Control to (1) the water balance maps not perfectly representing wa- Structure (see Harmar and Clifford, 2007). These same river mouth ter delivery to the coast and/or (2) the incorporation of the regions are used for all time steps in all models. internally drained Great Basin into the Columbia River and

Earth Surf. Dynam., 4, 831–869, 2016 www.earth-surf-dynam.net/4/831/2016/ A. D. Wickert: North American deglacial water routing 841

6 Table 1. Modern drainage basin areas and water (Qw) (cubic meters per second) and sediment (Qs) (10 metric tons per year) discharges. Many of these were gathered and independently of the Milliman and Farnsworth(2011) compilation, and generally provide good agreement with their numbers.

2 3 −1 −1 River A (km ) Qw (m s ) Qs (Mt yr ) Reference(s) Mackenzie 1.8 × 106 9910 128 Carson et al.(1998) Hudson Strait and 1.173 × 106 30 900a ∼1.1a,b Schneider-Vieira et al.(1994); Bentley(2006) Bay d Saint Lawrencec 1.344 × 106 ; 12 000d; 14 400e 3.6d Holeman(1968); Doyon(1996); Dolgopolova and e 1.61 × 106 Isupova(2011); Environment Canada(2013) Hudson River 0.034 × 106 590 0.737 Wall et al.(2008); Milliman and Farnsworth(2011) Susquehannaf 0.0704 × 106 1082 5 Gross et al.(1978); Kammerer(1990) Mississippig 3.225 × 106 18 430g 400h, 210j Kammerer(1990); Milliman and Farnsworth(2011) Rio Grande 0.870 × 106 570h,i; 22j 20h,i; 0.66j Gates et al.(2000); Milliman and Farnsworth (2011) Coloradok 0.637 × 106 ∼665h,i; 6.3j 145h,i, 0.1j US Bureau of Reclamation(1952); van Andel (1964) (as cited by Carriquiry and Sánchez, 1999); Kammerer(1990); Prairie and Callejo(2005); Nowak(2011); Milliman and Farnsworth(2011) Columbia 0.668 × 106 7500 7 Milliman and Meade(1983); Kammerer(1990)

a This is the sum of river inputs to Hudson Bay from pre-1975 data. b Modeled following Syvitski et al.(2003). c Not including modern diversions into or out of the Great Lakes (i.e., preindustrial discharge). These include the Chicago Diversion (out of the Great Lakes basin), and the Ogoki and Long Lake diversions (into the Great Lakes basin). The drainage basin ﬁgures show the modern Great Lakes–Saint Lawrence drainage basin, which includes these diversions; as these are small compared to other sources of possible error, I have not returned the mapped Great Lakes–Saint Lawrence drainage basin to its pre-diversion state. d The Great Lakes–Saint Lawrence drainage system alone (Dolgopolova and Isupova, 2011). e Including inﬂow into the Saint Lawrence estuary from rivers other than the Saint Lawrence f All drainages to Chesapeake Bay are included in the LGM calculations, even though only the Susquehanna is included at present (and in this table). This is because Chesapeake Bay is a ria, or drowned valley: the other rivers that drain into Chesapeake Bay joined the Susquehanna as tributaries at the LGM due to sea-level fall. g This is the sum of Mississippi River (Old River) discharge and drainage area, Mississippi River discharge and drainage area routed towards the Atchafalaya River, and the Atchafalaya–Red River system. Inclusion of the latter increases drainage area from 2.978 × 106 km3 to 3.225 × 106 km2 and discharge from 16 790 to 18 430 m3 s−1. h Before the installation of river control structures (i.e., preindustrial); all other values (including those on the same line) are modern unless indicated. i Before modern water withdrawals (i.e., preindustrial); these values are estimated based on the work of Gates et al.(2000) for the Rio Grande and Prairie and Callejo(2005) and Nowak(2011) for the Colorado River at Imperial Dam, which was summed with inputs from the Gila River (US Bureau of Reclamation, 1952); see Cohen et al.(2001) for an earlier calculation of natural discharge to the Colorado River delta and Schmidt et al.(2010) for a review of the 20th century human–river interaction that necessitated my approximation here. Modern values of water and sediment discharge are from Milliman and Farnsworth(2011). j Modern discharge to the ocean. k Modern discharge values are from Milliman and Farnsworth(2011); past (preindustrial) estimates and the estimated sediment yield are from other sources.

Colorado River basins because the flow-routing algorithm puted drainage basins and river discharge histories. I devel- (Metz et al., 2011) was set up to disallow internal drainage. oped these in two different ways: as paleogeographic maps of To provide more realistic discharge histories while also be- drainage divide positions, and as a simplified river discharge ing explicit about the raw outputs of the model, I rescaled history that defines periods of high and low flow. All radio- discharge in each basin to match its modern or preindustrial carbon ages included in these chronologies were recalibrated value with the mean of the computed time steps between 3 ka to calendar years BP using IntCal13 (Reimer et al., 2013). and present, while also making note on the plots (Figs. 11– I developed a set of data-driven drainage basin boundaries 16) of the predicted uncorrected modern or preindustrial for each of the study basins (Figs.7–10, lower right panels) discharge. This correction may create a systematic under- based on several lines of evidence. The first is the modern estimate of Columbia River discharge from 18.2 ± 0.3 ka drainage basin outlines of Lehner et al.(2013). The second +0.3 ± 14 (McGee et al., 2012) to 14.9−0.6 ka (12.6 0.15 C ka) is the ice-sheet margin chronology of Dyke et al.(2003) (Godsey et al., 2011; Reimer et al., 2013), when waters from and Dyke(2004), which, when lacking independent infor- Lake Bonneville – part of the Great Basin – flowed into the mation, I use as an approximate set of contours of ice-sheet Columbia River basin. thickness. The ice divides could, however, have moved with time, offsetting them from the chronological contours, and 3 Methods: data-driven drainage reconstruction Tarasov et al.(2012) and Gowan(2013) also note that uncertainties exist in this deglaciation chronology. Therefore, a A set of entirely data-driven drainage histories provides the key third set of information comes from the evidence of ice necessary counterpoint to evaluate the model-based com- streaming recently completed by Margold et al.(2014, 2015) www.earth-surf-dynam.net/4/831/2016/ Earth Surf. Dynam., 4, 831–869, 2016 842 A. D. Wickert: North American deglacial water routing

and the mapped Canadian eskers from Storrar et al.(2013). As there are few chronological controls on when these ice streams were active, ice divides are simply drawn between diverging ice streams. While this is a potential source of error, it also may not be such a bad assumption: where there are chronological controls on ice-stream reorganization, these have not been noted to change the continental- scale structure of ice divides (e.g., Stokes et al., 2009; Ross et al., 2009; Ó Cofaigh et al., 2010). Better chronologies often come from direct dating of ice-marginal positions (e.g., Licciardi et al., 1999; Dyke, 2004; Gowan, 2013), fluvial deposits (e.g., Bretz, 1969; Atwater, 1984; Ridge, 1997; Benito and O’Connor, 2003; Knox, 2007; Rittenour et al., 2007), and lacustrine deposits (e.g., Antevs, 1922; Kehew and Teller, 1994; Rayburn et al., 2005, 2007, 2011; Richard and Occhietti, 2005; Breckenridge, 2007; Breckenridge et al., 2012) and these provide the fourth layer to define drainage chronologies. The fifth layer of data comes from the offshore stratigraphic record (e.g., Leventer et al., 1982; Andrews and Tedesco, 1992; Andrews et al., 1999; Flower et al., 2004; Carlson et al., 2007a, 2009; Williams et al., 2012; Maccali et al., 2013), which can be compiled into paleohydrographs that may correlate with time-variable drainage basin area (Wickert et al., 2013), but outside of some information on provenance (Carlson et al., 2007a, 2009), it cannot independently indicate where the past drainage basin margins lie. While detailed and well-dated geological constraints per- mit a quantitative approach to reconstruct past discharge (Licciardi et al., 1999; Carlson et al., 2007a, 2009; Carl- son, 2009; Obbink et al., 2010; Wickert et al., 2013), most interpretations are much more basic, simply showing when periods of high or low discharge may have existed due to changes in drainage basin area (e.g., Ridge, 1997), isotopic ratios (e.g., Andrews and Dunhill, 2004), and/or proglacial lake outlets (e.g., Teller and Leverington, 2004; Brecken- ridge, 2015). A quantitative data–model comparison would be limited to a minority of both the total modeled river history and the available geologic data. I therefore took the sim- plest approach to interpret these records, defining them as simply “high discharge” (light and dark gray in Figs. 11– 16) or “low discharge” (white background in Figs. 11–16). Dark gray indicates times for which there is geological evidence for a distinct period high flow, and light gray indicates the time over which data-based inferences indicate that enhanced flow would have occurred and/or when discharge would have been higher but not as high as during the time indicated by the dark-gray regions. For example, Mississippi River discharge increased when Lake Agassiz flooded into it (one dark-gray band) but was overall higher than present Figure 6. Modeled sea-level curves at river mouths (Fig.5). River from the LGM until ∼ 12.9 ka due to ice-sheet input and a mouths formerly near (or under) thick ice have more variable sea- larger drainage basin area (light gray and dark gray). level histories, illustrating the divergent GIA histories result from different ice-sheet and solid-Earth models.

Earth Surf. Dynam., 4, 831–869, 2016 www.earth-surf-dynam.net/4/831/2016/ A. D. Wickert: North American deglacial water routing 843

Figure 7. End of the global LGM. North American drainage basins at the LGM. Ice models are organized left to right, top to bottom, in the order of their publication age. Drainage basins are delimited in black. Blue lines on the model output panels trace corridors through which total water discharge (ice melt and meteoric) is ≥ 1000 m3 s−1; these lines are based on raw outputs that are not adjusted based on modern or preindustrial discharge (Sect. 2.5, last paragraph). The data-driven compilation indicates evidence of past ice streams in blue (Margold et al., 2014). The model- and data-driven reconstructions show a generally consistent trend of an enlarged Mississippi River drainage basin, ice ﬂow via Hudson Strait into the North Atlantic, and ice-sheet meltwater routed down the Hudson and/or Susquehanna rivers.

4 Results tary video and in the data repository (Sect.7). The latter are a set of reconstructed deglacial water discharges by drainage The result of the computational work is two major prod- basin (Figs. 11–16), with Fig. 16 including the meltwater to- ucts: synthetic drainage basin and discharge histories. The tal discharge calculations of Licciardi et al.(1999) for com- former are summarized in four sets of maps showing signiﬁ- parison. These provide a spatially distributed and quantiﬁed cant points in time since the LGM (Figs.7–10), with the full time series of each reconstruction provided as a supplemen- www.earth-surf-dynam.net/4/831/2016/ Earth Surf. Dynam., 4, 831–869, 2016 844 A. D. Wickert: North American deglacial water routing

Figure 8. Bølling–Allerød warm period and (for all but ICE-3G and the data-driven reconstruction, for which the closest-available time slice(s) was/were chosen) Meltwater Pulse 1A. The Bølling–Allerød was the major period of warming between earlier near-LGM conditions and later Younger Dryas cooling. The early part of the Bølling–Allerød, the Bølling warming, encompasses MWP-1A, a period of 14–18 m sea-level rise between 14.65 and 14.31 ka (Deschamps et al., 2012). Collapse of the ice saddle between the Laurentide and Cordilleran ice sheets at this time, combined with generally enhanced melt rates due to the Bølling warming, could be the mechanism for rapid sea-level rise during MWP-1A (Gregoire et al., 2012, 2016; Gomez et al., 2015). Note the extension of the blue ≥ 1000 m3 s−1 river lines for many of the ice models; this is representative of an increase in melt that increases river discharge. view of how rivers in North America have broadly changed terpart” as opposed to “ground-truth” because these are also over the past 20 000 years. based on limited data and have varying degrees of quantita- The counterpart to these computed histories are those that tive rigor. Their key utility is that they offer a much simpler are driven by more direct interpretations of the data (past ex- path from the raw data to the interpreted drainage basin examples of this include those by Teller, 1990a; Licciardi et al., tents and periods of high discharge, and one that can be more 1999; Carlson et al., 2007a; Carlson, 2009). I write “coun- closely tied to geological fact. No age uncertainty is pro-

Earth Surf. Dynam., 4, 831–869, 2016 www.earth-surf-dynam.net/4/831/2016/ A. D. Wickert: North American deglacial water routing 845

Figure 9. Younger Dryas (at the transition to the Holocene for ICE-3G). The Younger Dryas was a period of rapid cooling that corresponds with meltwater rerouting away from the Mississippi (e.g., Williams et al., 2012; Wickert et al., 2013) and enhanced meltwater output to the Mackenzie (Murton et al., 2010; Keigwin and Driscoll, 2014) and Saint Lawrence (e.g., Carlson et al., 2007a; Cronin et al., 2012) rivers. Model predictions here vary widely, but all successful ice-sheet reconstructions should at least match the extensively researched continental- scale drainage rerouting away from the Mississippi Basin starting at 12.9 ka. Recent research indicates that outflow from Lake Agassiz, at the center of the continent, was first routed towards the Saint Lawrence and later (∼ 12.2 ka: Breckenridge, 2015) routed towards the Mackenzie River (Sect. 5.1.1). vided for these data-driven reconstructions: the maps display the large-scale drainage basins drawn in Figs.7–10, each multiple events that coincide, assembled from disparate data of which represents an important time in deglacial history, sources, of which not all have well-quantified error. The pe- display the current state of knowledge for that point in the riods of high and low discharge are marked with sharp lines geologic timescale, regardless of chronological error. There- even where chronological constraints are sparse, and should fore, large-scale disagreements between the data-constrained be viewed as a current geological best estimate. Nevertheless, drainage basin outlines and the model results at these key www.earth-surf-dynam.net/4/831/2016/ Earth Surf. Dynam., 4, 831–869, 2016 846 A. D. Wickert: North American deglacial water routing

Figure 10. Early Holocene. The early Holocene was notable for its ice retreat and proglacial lake growth, which included enhanced flow down the Saint Lawrence River and abrupt drainage of proglacial lakes Agassiz and Ojibway into Hudson Bay. In G12, westerly flow around the ice dam in Hudson Bay could highlight a pathway for the 8.6 ka partial drainage of Lake Agassiz–Ojibway (Breckenridge et al., 2012). Many other rivers no longer had ice-sheet input. In ICE-5G, the upper Missouri river occupies the preglacial Heart River valley in the Bismarck–Mandan area (Todd, 1914; Leonard, 1916; Bluemle, 2006) and flows into the James River, tributary to the Red River of the north, and away from the Mississippi (Sect. 5.1.3). This illustrates how sensitive low-gradient mid-continental river systems can be to subtle tilts of the land surface, especially when preexisting valleys and topographic slopes facilitate alternate flow paths. points in the geological timescale generally indicate a need (Sect.3, last paragraph). These potential mismatches are dis- to improve the model-based reconstructions. Mismatches be- cussed below. tween data-driven and model-driven discharge histories are less straightforward to attribute to model error, as they involve both drainage basin area and rates of ice-sheet melt

Earth Surf. Dynam., 4, 831–869, 2016 www.earth-surf-dynam.net/4/831/2016/ A. D. Wickert: North American deglacial water routing 847

5 Discussion the ice again retreated during the 15.9–15.6 ka Mackinaw Interstade, the Hudson River captured outflow from all of 5.1 Comparison with geologic evidence for varying the Great Lakes except Superior, reducing Mississippi dis- drainage basin areas and discharge charge and removing ice-sheet inputs from the Susquehanna for the final time. Ice advanced again starting at 15.6 ka to- Each of the river systems in this study experienced high wards its Port Huron Stade maximum, at which point only flow and low flow at different times. High flow could be Lake Ontario flowed towards the Hudson River (via the Mo- due to meltwater inputs, drainage area expansion, or an in- hawk Valley) and all other Great Lakes sent their outflow − crease in P ET. The first two are often linked: ice-sheet to the Mississippi (Hansel and Mickelson, 1988; Licciardi advance into mid-continent drainage basins vastly increased et al., 1999; Obbink et al., 2010). At 14.2 ka, ice retreat dur- their areas (Figs.7–9). Figure 11 includes a brief description ing the Two Creeks Interstade rerouted outflow from all of of each of these times, which is expanded upon in the fol- the Great Lakes except for the western Lake Superior basin to lowing sections. The data-driven drainage histories (Sect.3) the Hudson River (Broecker and Farrand, 1963; Ridge, 1997; through North America are updated from Wickert(2014), Licciardi et al., 1999; Rech et al., 2012), though the Missis- with the dates for ice advance and retreat compiled by Lic- sippi was augmented by increasing melt via the Des Moines ciardi et al.(1999) still broadly consistent with the modern Lobe, James Lobe, and Lake Agassiz (Patterson, 1997, 1998; state of knowledge. For the following chronology, all radio- Licciardi et al., 1999; Obbink et al., 2010; Sionneau et al., carbon ages were calibrated (or recalibrated) using IntCal13 2010). Outflow from Lake Michigan temporarily returned to (Reimer et al., 2013). Current geologic evidence, while in- the Mississippi during the Greatlakean Stade, 13.5–13.0 ka complete and subject to revision, provides a good picture of (Evenson et al., 1976; Hansel and Mickelson, 1988; Liccia- the broad changes in drainage across North America since rdi et al., 1999) before rapid and complete (though not final) the LGM. abandonment of the Mississippi River meltwater outlet between 13 and 12.85 ka. 5.1.1 Drainage histories by river Much of this early history stems from the continental record of ice advance and retreat (e.g., Dreimanis and The Mississippi, Hudson, and Susquehanna rivers traded Goldthwait, 1973). Thus, it is a good predictor of changes drainage inputs from the Great Lakes and their respective for- in drainage basin area, but is somewhat less effective at mer ice lobes, resulting in a pattern of symmetric increases predicting past river discharge, especially when that dis- and decreases in discharge as ice advanced and retreated charge is strongly influenced by ice-sheet melt. Discharge along the southern Laurentide margin. At the LGM, the inferences continue to be aided by the growing database drainage area of the Great Lakes was split between the Mis- of paleoceanographic proxy records (e.g., Keigwin et al., sissippi and Susquehanna rivers (Flock, 1983; Ridge et al., 2006; Obbink et al., 2010; Williams et al., 2012; Hillaire- 1991; Ridge, 1997; Licciardi et al., 1999; Reusser et al., Marcel et al., 2013; Taylor et al., 2014; Gibb et al., 2014; 2006; Knox, 2007), with the Mississippi receiving meltwa- Gil et al., 2015), improved geochemical techniques to pro- ter from all but a portion of Lake Ontario. The Hudson River duce richer records that also reduce or quantify uncertainties received some northern ice-sheet inputs (Wickert, 2014) as (e.g., Vetter, 2013; Spero et al., 2015; Khider et al., 2015; well as flow from the Delaware River due to glacial isostatic Vázquez Riveiros et al., 2016), and continuing work to pro- tilting of the land surface (Stanford et al., 2016). At 19.9 ka, vide quantitative freshwater discharge estimates based on ice retreat associated with the Erie Interstade (Mörner and these data (Carlson et al., 2007a; Carlson, 2009; Obbink Dreimanis, 1973; Monaghan, 1990) opened the Mohawk val- et al., 2010; Wickert et al., 2013). Particularly important ley in New York, which became the preferential pathway are records of past water isotopes, past water temperatures, for all of the meltwater formerly sent to the Susquehanna and geochemical signatures of provenance. In addition to River, as well as water from the Lake Erie and western these sediment-based methods, continental-scale constraints Lake Ontario basins that formerly drained to the Mississippi on ice-sheet melt can be informed by sea-level fingerprint- (Ridge, 1997). In spite of the loss of Mississippi-routed melting (e.g., Gomez et al., 2015; Liu et al., 2015). In this work, water inputs, deposits of the Kankakee Outwash Plain (Knox, I follow Licciardi et al.(1999) in correlating the terrestrial 2007) indicate increased meltwater outflow to the Missis- record of ice-sheet advance through the Great Lakes with a sippi River from the Lake Michigan basin ∼ 19.3–18.8 ka, large eastern Mississippi River drainage basin and therefore with a major flood ∼ 19 ka (Curry et al., 2014). Following the with periods of enhanced Mississippi River discharge. end of the Erie Interstade, at 18.4 ka (Licciardi et al., 1999; Meltwater discharge down the Mississippi decreased Reimer et al., 2013; Wickert, 2014), the Port Bruce Stade ad- sharply between 13.0 and 12.85 ka (Flower et al., 2004; vance produced the Valley Heads glaciation in New York, Williams et al., 2012; Wickert et al., 2013), synchronous sending flow back from the Hudson River to the Susque- with ice retreat that opened the Saint Lawrence River (Oc- hanna and Mississippi from 18.4 to 15.9 ka (see Reusser chietti and Richard, 2003; Richard and Occhietti, 2005; et al., 2004, 2006, for the Susquehanna terrace record). As Carlson et al., 2007a; Rayburn et al., 2011). At 12.9 ka www.earth-surf-dynam.net/4/831/2016/ Earth Surf. Dynam., 4, 831–869, 2016 848 A. D. Wickert: North American deglacial water routing

Figure 11. Computed discharge from ICE-3G/VM1 (Tushingham and Peltier, 1991), with gray shading at times of enhanced discharge based on the geologic record. Dark gray indicates a specific time of enhanced discharge, while light gray indicates inferred higher discharge from ice-sheet and landscape geometry or other factors, or, for the Columbia River, a time of enhanced discharge that is likely less-enhanced than discharge earlier in the record. These are annotated here with the following abbreviations: LIS – Laurentide Ice Sheet; H0, H1B, and H1A – Heinrich events 0, 1B, and 1A; CIS – Cordilleran Ice Sheet. the Lake Agassiz drainage subbasin was rerouted away 2016), though this rerouting did not produce any significant from the Mississippi River. The best current chronol- drawdown of the lake level. Starting at 12.18 ± 0.48 ka, out- ogy indicates that water from the Lake Agassiz subbasin flow from the Lake Agassiz subbasin was rerouted to the first flowed east to the newly ice-free Saint Lawrence northwest towards the Mackenzie River (Andrews and Dun- (Broecker et al., 1989; Carlson et al., 2007a; Carlson and hill, 2004; Carlson et al., 2007a; Breckenridge, 2015). Two Clark, 2012; Breckenridge, 2015; Levac et al., 2015; Leydet, final periods of meltwater discharge down the Mississippi

Earth Surf. Dynam., 4, 831–869, 2016 www.earth-surf-dynam.net/4/831/2016/ A. D. Wickert: North American deglacial water routing 849

Mississippi River Susquehanna/Chesapeake 160000 40000 140000 35000 120000 30000 100000 25000 80000 20000 60000 15000 40000 10000 20000 5000 0 0 0 5 10 15 20 0 5 10 15 20

Hudson River Saint Lawrence River 160000 160000 140000 140000 120000 120000 100000 100000 80000 80000 60000 60000 40000 40000 20000 20000 0 0 0 5 10 15 20 0 5 10 15 20 ]

1 Hudson Strait Mackenzie River 250000 160000 −

s 140000

3 200000 120000 100000 m 150000

[ 80000

100000 60000 e 40000 g 50000

r 20000

a 0 0 0 5 10 15 20 0 5 10 15 20 h c

s Columbia River Colorado River

i 18000 2000 16000 D 14000 1500 12000 10000 1000 8000 6000 4000 500 2000 0 0 0 5 10 15 20 0 5 10 15 20

Rio Grande 1200 1000 Total discharge 800 Meltwater discharge 600 Meteoric water (P-ET) discharge 400 Uncorrected modern 200 modeled discharge 0 0 5 10 15 20 Age [ka]

Figure 12. Computed discharge based on the ANU ice-sheet–solid-Earth model pair (Lambeck et al., 2002), compared to the geologic record of enhanced past discharge (gray shading: dark, direct record; light, inferred and/or less-enhanced discharge). Notes on the geologic record can be found in Fig. 11.

River occurred after the initial rerouting of Lake Agassiz Superior basin back towards the Mississippi River. The sec- flow. The first is recorded in a 11.8–11.3 ka δ18O spike in ond lasted ∼ 11.0–10.6 ka (Fisher, 2003; Wickert, 2014), and offshore sediments (Williams et al., 2012; Wickert, 2014), was the final pulse of ice-sheet meltwater to flow down the whose age range encompasses the ∼ 11.6–11.5 ka Marquette Mississippi. Readvance of ice in the Lake Superior basin (Lowell et al., The Saint Lawrence continued to receive significant ice- 1999) that temporarily separated the Great Lakes–Saint sheet meltwater inputs via proglacial lakes until some- Lawrence river system from Lake Agassiz (Breckenridge and time between ∼8.6 ka, when the combined Lake Agassiz– Johnson, 2009) and may have sent meltwater from the Lake Ojibway level dropped in a partial drainage event (Brecken-

www.earth-surf-dynam.net/4/831/2016/ Earth Surf. Dynam., 4, 831–869, 2016 850 A. D. Wickert: North American deglacial water routing

Mississippi River Susquehanna/Chesapeake 160000 40000 140000 35000 120000 30000 100000 25000 80000 20000 60000 15000 40000 10000 20000 5000 0 0 0 5 10 15 20 0 5 10 15 20

Hudson River Saint Lawrence River 160000 160000 140000 140000 120000 120000 100000 100000 80000 80000 60000 60000 40000 40000 20000 20000 0 0 0 5 10 15 20 0 5 10 15 20 ]

1 Hudson Strait Mackenzie River 250000 160000 −

s 140000

3 200000 120000 100000 m 150000

[ 80000

100000 60000 e 40000 g 50000

r 20000

a 0 0 0 5 10 15 20 0 5 10 15 20 h c

s Columbia River Colorado River

i 18000 2000 16000 D 14000 1500 12000 10000 1000 8000 6000 4000 500 2000 0 0 0 5 10 15 20 0 5 10 15 20

Rio Grande 1200 1000 Total discharge 800 Meltwater discharge 600 Meteoric water (P-ET) discharge 400 Uncorrected modern 200 modeled discharge 0 0 5 10 15 20 Age [ka]

Figure 13. Computed discharge based on ICE-5G/VM2 (Peltier, 2004), compared to the geologic record of enhanced past discharge (gray shading: dark, direct record; light, inferred and/or less-enhanced discharge). Notes on the geologic record can be found in Fig. 11. ridge et al., 2012), and 8.2 ka, when the ice saddle in Hudson River and into the Gulf of Saint Lawrence until ∼ 7.8 ka (Oc- Bay collapsed, causing a massive flood that resulted in sea- chietti et al., 2004). surface freshening and the brief cooling of the 8.2 ka event The record of ice and meltwater discharge from Hudson (Barber et al., 1999; Carlson et al., 2009; Roy et al., 2011; Bay and Hudson Strait that predates the Agassiz–Ojibway Hoffman et al., 2012; Gregoire et al., 2012; Breckenridge flood comes primarily from ice-rafted debris in marine sedi- et al., 2012; Stroup et al., 2013). While this ended melt- ment cores. This is because Hudson Strait was an ice-covered water inputs to the Saint Lawrence River, meltwater from marine-terminating margin of a major ice stream of the Lau- the Québec–Labrador dome flowed down the Manicouagan rentide Ice Sheet (Margold et al., 2014). The largest of the post-LGM ice-rafted debris events was Heinrich Event 1,

Earth Surf. Dynam., 4, 831–869, 2016 www.earth-surf-dynam.net/4/831/2016/ A. D. Wickert: North American deglacial water routing 851

Mississippi River Susquehanna/Chesapeake 160000 40000 140000 35000 120000 30000 100000 25000 80000 20000 60000 15000 40000 10000 20000 5000 0 0 0 5 10 15 20 0 5 10 15 20

Hudson River Saint Lawrence River 160000 160000 140000 140000 120000 120000 100000 100000 80000 80000 60000 60000 40000 40000 20000 20000 0 0 0 5 10 15 20 0 5 10 15 20 ]

1 Hudson Strait Mackenzie River 250000 160000 −

s 140000

3 200000 120000 100000 m 150000

[ 80000

100000 60000 e 40000 g 50000

r 20000

a 0 0 0 5 10 15 20 0 5 10 15 20 h c

s Columbia River Colorado River

i 18000 2000 16000 D 14000 1500 12000 10000 1000 8000 6000 4000 500 2000 0 0 0 5 10 15 20 0 5 10 15 20

Rio Grande 1200 1000 Total discharge 800 Meltwater discharge 600 Meteoric water (P-ET) discharge 400 Uncorrected modern 200 modeled discharge 0 0 5 10 15 20 Age [ka]

Figure 14. Computed discharge based on the G12/VM2 coupled ice-sheet–solid-Earth model (Gregoire et al., 2012; Peltier, 2004), compared to the geologic record of enhanced past discharge (gray shading: dark, direct record; light, inferred and/or less-enhanced discharge). Notes on the geologic record can be found in Fig. 11. which occurred in two pulses: H1A (16.8–16.3 ka) and H1B Miller, 1993; Metz et al., 2008; Rashid et al., 2014; Pearce, (15.9–15.7 ka) (Gil et al., 2015). Heinrich Event 0, formerly 2015; Jennings et al., 2015); these comprise the 11.0–10.6 ka thought to have an age of 12.8–12.3 ka, which corresponds Gold Cove advance (Pearce, 2015) and the 9.7–9.2 ka Noble to the Younger Dryas (Andrews and Tedesco, 1992; Clark Inlet advance (Jennings et al., 2015). Following these ice- et al., 2001), has been recently shown to correlate with early rafting events, the merged Lake Agassiz–Ojibway released Holocene warming and melt from 11.5–11.3 ka (Pearce et al., water into Hudson Bay in two major pulses. The ﬁrst was the 2015). Later ice-rafting events are related to ice advances of ∼ 8.6 ka partial drainage of Lake Agassiz–Ojibway, identi- the Québec–Labrador dome into Ungava Bay (Kaufman and ﬁed by a low-water period in the varve record (Breckenridge

www.earth-surf-dynam.net/4/831/2016/ Earth Surf. Dynam., 4, 831–869, 2016 852 A. D. Wickert: North American deglacial water routing

Mississippi River Susquehanna/Chesapeake 160000 40000 140000 35000 120000 30000 100000 25000 80000 20000 60000 15000 40000 10000 20000 5000 0 0 0 5 10 15 20 0 5 10 15 20

Hudson River Saint Lawrence River 160000 160000 140000 140000 120000 120000 100000 100000 80000 80000 60000 60000 40000 40000 20000 20000 0 0 0 5 10 15 20 0 5 10 15 20 ]

1 Hudson Strait Mackenzie River 250000 160000 −

s 140000

3 200000 120000 100000 m 150000

[ 80000

100000 60000 e 40000 g 50000

r 20000

a 0 0 0 5 10 15 20 0 5 10 15 20 h c

s Columbia River Colorado River

i 18000 2000 16000 D 14000 1500 12000 10000 1000 8000 6000 4000 500 2000 0 0 0 5 10 15 20 0 5 10 15 20

Rio Grande 1200 1000 Total discharge 800 Meltwater discharge 600 Meteoric water (P-ET) discharge 400 Uncorrected modern 200 modeled discharge 0 0 5 10 15 20 Age [ka]

Figure 15. Computed discharge based on the ICE-6G/VM5a coupled ice-sheet–solid-Earth model (Argus et al., 2014; Peltier et al., 2015), compared to the geologic record of enhanced past discharge (gray shading: dark, direct record; light, inferred and/or less-enhanced discharge). Notes on the geologic record can be found in Fig. 11. et al., 2012). The second was the 8.2 ka complete drainage events, from the start of the Noble Inlet Advance at 9.7 ka of Lake Agassiz–Ojibway (Barber et al., 1999; Brecken- until ∼ 8.0 ka (Jennings et al., 2015). ridge et al., 2012; Stroup et al., 2013) that is associated with After 8.2 ka, Hudson Bay continued to receive a smaller collapse of the Laurentide Ice Sheet saddle that connected meltwater input from the ablating remnant Laurentide ice the Keewatin and Québec–Labrador domes (Gregoire et al., caps (Carlson et al., 2008) that included one last ∼ 200-year 2012). High carbonate concentrations and low δ18O in sedi- period of enhanced discharge ∼ 7.55–7.35 ka that appears in ments indicate consistent high discharge, even between these the δ18O record of Jennings et al.(2015). The total volume of the 7.5–6.8 ka ﬁnal ablation of the Laurentide Ice Sheet

Earth Surf. Dynam., 4, 831–869, 2016 www.earth-surf-dynam.net/4/831/2016/ A. D. Wickert: North American deglacial water routing 853

Mississippi River Susquehanna/Chesapeake 160 000 40 000 140 000 35 000 120 000 30 000 100 000 25 000 80 000 20 000 60000 15 000 4 0 000 10 000 20 000 5000 1 0 0 − 0 5 10 15 20 0 5 10 15 20 3 Hudson River Saint Lawrence River 160 000 160 000 140 000 140 000 120 000 120 000 100 000 100 000 80 000 80 000 60 000 60 000 40 000 40 000 20 000 20 000 0 0 0 5 10 15 20 0 5 10 15 20

Hudson Strait Mackenzie River 250 000 160 000 140 000 200 000 120 000 150 000 100 000 80 000 100 000 60 000 40 000 50 000 20 000 0 0 0 5 10 15 20 0 5 10 15 20

Columbia River Colorado River 18 000 2000 16 000 14 000 1500 12 000 10 000 1000 8000 6000 4000 500 2000 0 0 0 5 10 15 20 0 5 10 15 20

Rio Grande 1200 1000 ICE-3G/VM1 ICE-5G/VM2 800 ICE-6G/VM5a 600 Total discharge (meltwater + meteoric water) [mwater) meteoric + (meltwater s discharge Total ] ANU 400 G12/VM2 200 Licciardi et al. (1999) 0 0 5 10 15 20 Age [ka]

Figure 16. Total discharge from each river basin, represented on a single set of axes to compare the timing and absolute magnitudes of meltwater discharge among the models, with the geologic record of enhanced past discharge (gray shading: dark, direct record; light, inferred and/or less-enhanced discharge). Also shown here in darker gray lines on four panels are the data-driven discharge histories of Licciardi et al. (1999), with their original radiocarbon chronologies calibrated using IntCal13 (Reimer et al., 2013).

– comprising the Keewatin, Québec–Labrador, and Baffin charge (Table1). Therefore, background melt outside of the Island–Foxe ice domes – could have been up to 5 m sea- δ18O excursion does not appear as a time of enhanced dis- level equivalent (SLE) (Carlson et al., 2007b; Jennings et al., charge in Figs. 11–16. Following final Laurentide Ice Sheet 2015). Much of this melt would have flowed into Hudson collapse, isostatic rebound with a time-averaged uplift rate Bay, but would not have been significant: the maximum es- of 1.3 cm yr−1 (Lavoie et al., 2012) caused Hudson Bay to timate (5 m SLE and all melt entering Hudson Bay) results shrink to its modern size. in an average discharge enhancement of 82 m3 s−1, which The main period of post-LGM high discharge down the is small compared to the 30 900 m3 s−1 total modern dis- Mackenzie River started ∼ 15.6 ka with the retreat of the www.earth-surf-dynam.net/4/831/2016/ Earth Surf. Dynam., 4, 831–869, 2016 854 A. D. Wickert: North American deglacial water routing

Mackenzie Lobe of the Laurentide Ice Sheet and associ- Wright, 1985; He, 2011; Oster et al., 2015; Oster and Kel- ated meltwater production (Duk-Rodkin et al., 1994; Lem- ley, 2016). These cool, wet periods are marked by high plu- men et al., 1994). Meltwater inputs increased at 13.6 ± 0.2 ka vial lake levels (e.g., Currey et al., 1990; Matsubara and (Smith, 1994), when glacial Lake McConnell formed in the Howard, 2009; McGee et al., 2012; Oviatt, 2015), mountain Great Bear Lake Basin. Lake McConnell catastrophically glacier advances prior to the onset of the Bølling–Allerød flooded towards the Mackenzie River between ∼ 13.6 and (e.g., Owen et al., 2003; Guido et al., 2007; Orme, 2008; Ref- ∼ 13.3 ka, possibly again at ∼ 13.3 ka, and at ∼ 13.1 ka. At snider et al., 2008, 2009; Brugger, 2010), increased landslid- ∼ 13.1 ka, ice retreat rerouted glacial Lake Peace, which for- ing in the Rio Grande basin between ∼ 21.2 and ∼ 14.5 ka merly drained into the Missouri River (and hence the Mis- (Reneau and Dethier, 1996), deposition of valley fills in sissippi), towards the Mackenzie, significantly increasing the the Colorado River (Pederson et al., 2013), and speleothem drainage area of the latter. At 12.18 ± 0.48 ka, strandlines of growth (e.g., Polyak et al., 2004; Oster and Kelley, 2016). glacial Lake Agassiz indicate that it, too, started to flow to The warmer Bølling–Allerød and Holocene correspond to the north (Breckenridge, 2015). This age range is consis- records of mountain glacier retreat (Guido et al., 2007; Laabs tent with the 13.0 ± 0.2 to 11.7 ± 0.1 bounds on the age of et al., 2013; Munroe and Laabs, 2013) and strath terrace for- gravel deposits that overlie a regional erosion surface on the mation due to river incision (Anders et al., 2005; Cook et al., Mackenzie River delta, presumably the result of a large flood 2009). While anomalous records, such as speleothem growth associated with drainage rearrangement (Murton et al., 2010; ∼ 14 ka during the Bølling–Allerød (Polyak et al., 2004), Carlson and Clark, 2012). It also postdates the youngest complicate the picture of past climate in the southwest, the dated material from glacial Lake Mackenzie (13.0 ± 0.3 ka), general pattern of climate change closely mirrors that ob- which drained when the Ramparts – the largest bedrock served in the Greenland ice core records (Benson et al., 1997; gorge in the Mackenzie River – incised (Mackay and Math- Svensson et al., 2008; Asmerom et al., 2010; Wagner et al., ews, 1973). This incision may be related to drainage rerout- 2010). ing that increased inflow into glacial Lake Mackenzie and caused it to overtop the bedrock sill that dammed it to the 5.1.2 Comparison of data and models for river north. Meltwater inputs to the Mackenzie River ended no discharge later than 11 ka, when its eastern tributaries were temporarily rerouted eastward due to a combination of ice retreat and In order to determine how well, broadly, these different mod- glacial isostatic depression (see Sect. 5.1.3). Later northwest- els were able to reproduce our observed record of discharge ern drainage of Lake Agassiz during its high-water Emer- variability, I subdivided the computed discharges into events son Phase, 10.8–10.2 ka (Fisher, 2007), must therefore have during known high-flow periods (light and dark-gray re- flowed in a northerly course near or along the retreating gions in Figs. 11–16) and those from the low-flow remainder ice margin instead of entering the isostatically constricted of the record. I first compared the mean discharges during Mackenzie River drainage basin. known high- and low-flow periods (Fig. 17a). For the five The Columbia River has an offshore record of rapid tur- ice-sheet models and nine river systems, only three times bidite deposition that stretches 36.1–10.6 ka (Zuffa et al., did the known low-flow periods correspond to an overall 2000). Within this time span, large floods associated with higher computed discharge: the Hudson River for the ANU glacial Lake Missoula occurred 22.9–16.5 ka (Bretz, 1923, model and Hudson Strait for ICE-3G and ICE-5G. The ice- 1969; Waitt, 1980; Atwater, 1984; Benito and O’Connor, physics-based G12 performed the best in generating dis- 2003; Lopes and Mix, 2009). Lake Bonneville routed a por- charges during known high-flow periods that were much tion of flow from the Great Basin to the Columbia River from higher than those during known low-flow periods, with ICE- 18.2 ± 0.3 to 16.4 ± 0.2 ka (Gilbert, 1890; Godsey et al., 6G in second place and ICE-3G and ICE-5G tied for third. 2005, 2011; McGee et al., 2012), including a massive < 1- The strong performance of G12 may be because its input year flood at ∼ 17.5 or ∼ 18 ka that resulted from rapid mass balance was driven by a coupled ocean–atmosphere ∼ 100 m incision of a natural dam of alluvium (Jarrett and GCM (FAMOUS: Smith et al., 2008), making its ice mass Malde, 1987; Godsey et al., 2005, 2011; Oviatt, 2015). The the result of a precomputed hydrologic balance. I then gen- terrestrial record contains evidence for later, smaller outflows erated histograms for the high-flow and low-flow segments from glacial Lake Missoula 15.1–12.6 ka (Hanson et al., of the model results and performed Kolmogorov–Smirnov 2012), and intermittent overflow inputs from Lake Bon- tests between the pairs of distributions for each ice model ± +0.3 neville from 16.4 0.2 to 14.9−0.6 ka (synthesis of Godsey and river system to test how different they are from one an- et al., 2011; McGee et al., 2012). other (Fig. 17b). Here, the GIA-based ICE-6G and ICE-5G Water discharge in Colorado River and the Rio Grande, were second and third, respectively, to ICE-3G, though ICE- while influenced by mountain glacier retreat, responded pri- 3G cannot be a correct ice-sheet reconstruction more gener- marily to the cooler and wetter regional climate during ally because it does not include a large enough total ice mass. the LGM and Younger Dryas that was caused by changes G12 reproduced many patterns of deglacial water discharge in continental-scale moisture transport (e.g., Kutzbach and well, including the pulses of iceberg/meltwater output from

Earth Surf. Dynam., 4, 831–869, 2016 www.earth-surf-dynam.net/4/831/2016/ A. D. Wickert: North American deglacial water routing 855

tent time steps between ice-sheet reconstructions can affect (a) the fits, as can be seen especially for the Rio Grande and Columbia River, which should be the same for all runs as the same climate reconstruction was employed in all cases. Therefore, these catchments were removed from the above analysis, the broad strokes of which provide a useful template for comparison. Across all model runs, Hudson Strait was the consistent site of the worst agreement between data and the models. It is possible that this is due to data-collection issues (e.g., few organics) and data sparsity (few settlements and fewer GPS monuments) in this formerly fully ice-covered region, leading to a lack of good regional constraints on ice-sheet thickness near the center of the Laurentide Ice Sheet. How- ever, recent data sets have brought more clarity to the Hud- (b) son Strait record (e.g., Rashid et al., 2014; Gil et al., 2015; Pearce et al., 2015; Pearce, 2015; Jennings et al., 2015), fur- thering the argument that this mismatch could be because the models do not accurately represent the dynamics of the marine-terminating Hudson Strait ice catchment (e.g., MacAyeal, 1993). An additional complication of Hudson Bay/Strait is that its drainage area was smaller at the LGM (Fig.7), but it had the potential to release a large amount of ice, as is observed in the Heinrich events (e.g., Heinrich, 1988). A final potential issue is that Hudson Bay lies on the low-relief Canadian Shield in the zone of maximum glacial isostatic deformation, meaning that its computed drainage area during deglaciation is sensitive to GIA (Fig. 10). In sum- mary, a proper comparison requires both long-term proxy discharge evidence and a mechanistic ice-sheet and GIA Figure 17. Computed discharges are classified as occurring at times model that can explain the dynamics of ice and solid-Earth of known high or low flow based on geologic record. Times of masses in Hudson Bay and Hudson Strait under a realistic cli- high flow in the geologic record are denoted by both light and dark mate forcing. The Glimmer–CISM (Rutt et al., 2009) runs of gray in Figs. 11–16. (a) Normalized difference in the modeled dis- Gregoire et al.(2012) are able to reconstruct the general form charge between geologically known high and low discharge times. of the pulsed meltwater releases seen in proxies (e.g., Jen- The greatest mean normalized difference is seen in G12, the ice- nings et al., 2015), putting more weight behind approaches physics-based model of Gregoire et al.(2012), and the least in the GIA-based model of Lambeck et al.(2002). (b) The Kolmogorov– that involve ice physics to reconstruct past ice sheets (e.g., Smirnov test measures the distance between one-dimensional dis- Tarasov and Peltier, 2005, 2006; Tarasov et al., 2012; Gre- tributions; here I use it to compute a metric of difference between goire et al., 2012; Wickert et al., 2013). the computed discharge histograms for each river system during the A comparison between these model results and data-driven generalized data-derived times of high and low discharge. A high drainage histories of Licciardi et al.(1999) (Fig. 16: Mis- distance value indicates a good match between the data and model. sissippi, Hudson, Saint Lawrence, Hudson Strait) includes Here, the GIA-based ICE-3G and ICE-6G models perform the best, both similarities and differences when compared to the river with poor matches only to Hudson Bay. The ice-physics-based G12 system reconstructions presented here. For the Mississippi performs poorly, due in part to the too early demise of its Québec– River System, Licciardi et al.(1999) expect enhanced dis- Labrador dome, and the GIA-based ANU model performs the worst. charge between 18.4 and 16.3 ka and prior to 19.9 ka – that is, at times when all Great Lakes flow was routed down the Mississippi River. The reconstructions analyzed here exhibit Hudson Strait (Fig. 14), but its goodness of fit is reduced by enhanced discharge over the entire time prior to 16.3 ka, but its short (but intense) pulse of meltwater down the Macken- with no shorter-term periods of more enhanced discharge zie River, its lack of early meltwater delivery to the Hud- (Fig. 16: Mississippi). The discharges predicted by Liccia- son River, and a too early switch of meltwater outflow away rdi et al.(1999) are higher than any produced here, and this from the Mississippi River because its climate forcing and is because they neglect evapotranspiration. Periodicity due to dynamical ice response caused the Québec–Labrador dome drainage rearrangement, predicted by Licciardi et al.(1999), to retreat too early. For both of these metrics, the inconsis- is seen in the δ18O record of the Gulf of Mexico (Williams www.earth-surf-dynam.net/4/831/2016/ Earth Surf. Dynam., 4, 831–869, 2016 856 A. D. Wickert: North American deglacial water routing et al., 2012; Wickert, 2014) (albeit with somewhat different routing. This is the result of its ice sheet briefly crossing a timing, especially earlier in the record), though this is more threshold that rerouted drainage. It is these thresholds that sensitive to the very negative δ18O of meltwater discharge make drainage routing an especially useful discriminator to than to the overall freshwater discharge. The modeled Hud- separate valid and invalid portions of ice-sheet reconstruc- son River discharge does not exhibit the predicted Erie or tions. Mackinaw Interstade periods of enhanced discharge in any During Bølling–Allerød time, only ICE-6G and G12 prop- of the computed discharge histories presented here. Com- erly show that ice retreated beyond the bounds of the Susque- pelling geological evidence that meltwater flowed through hanna River basin. All models simulate the Mississippi and the Mohawk Valley into the Hudson River during the Erie Mackenzie catchments generally well, and close to what Interstade (Ridge, 1997) makes this omission a point to im- available data (e.g., Margold et al., 2014) would indicate. prove in future reconstructions that include the southern and During the early part of the Younger Dryas, the Macken- southeastern Laurentide Ice Sheet. The Saint Lawrence River zie and Saint Lawrence rivers both experienced periods constraints from Licciardi et al.(1999) are consistent with of high discharge (de Vernal et al., 1996; Carlson et al., the flow routing of G12 (Gregoire et al., 2012) and ICE- 2007a; Cronin et al., 2012; Murton et al., 2010; Keigwin and 6G (Argus et al., 2014; Peltier et al., 2015), even though Driscoll, 2014). Debate continues as to whether Lake Agas- G12 reroutes meltwater discharge from the Mississippi to siz flowed towards the Mackenzie River in the northwest the Saint Lawrence ∼ 1.1 ka too early (Wickert et al., 2013). (e.g., Murton et al., 2010; Keigwin and Driscoll, 2014), to- The Hudson Strait Record is difficult to match because of wards the Saint Lawrence River in the east (e.g., Carlson and its many episodic meltwater releases (e.g., Heinrich, 1988; Clark, 2012), or, as one group has proposed, became a closed Andrews and MacLean, 2003; Hemming, 2004; Gil et al., basin at this time due to regional drying (Lowell et al., 2013). 2015; Jennings et al., 2015), and only the climate-forced However, evidence currently indicates that Lake Agassiz and ice-physics-based G12 reproduces the general form of drained east at the onset of the Younger Dryas (Carlson et al., these events. These events are simplified or smoothed-over 2007a; Rayburn et al., 2011; Carlson and Clark, 2012; Ley- by Licciardi et al.(1999), but recent improvements in the age det, 2016), and then drained north at 12.18 ± 0.48 ka (Breck- constraints on these iceberg and/or meltwater release events enridge, 2015), which is consistent with the data of Fisher (e.g., Rashid et al., 2014; Gil et al., 2015; Pearce et al., 2015; et al.(2008) and Murton et al.(2010). The drying hypothe- Pearce, 2015; Jennings et al., 2015) set the stage for forth- sis seems the least likely, as the model to simulate it (Lowell coming advances in reconstructions of the northeastern core et al., 2013) creates a closed basin when at least two param- of the Laurentide Ice Sheet. eters seem unreasonable. These parameters are (1) evapora- tive loss that is generally > 1.5 times today’s observations, 5.1.3 Comparison of data and models for drainage in spite of cooler temperatures and a longer lake-ice sea- basin extents son, although katabatic winds may have increased evaporation somewhat over the open water (see Carlson and Clark, Any model-based reconstruction that employs a realistically 2009, for more on the problem of too much evaporation), and shaped North American ice-sheet complex will produce cer- (2) a temperature lapse rate that is 7.3 ◦C km−1, outside of the tain changes in continental-scale drainage. The most visible error bars of the 5.1 ± 1.2 ◦C km−1 observed on modern ice such pattern, consistent across all model-based reconstruc- sheets and ice caps Anderson et al.(2014), and resulting in a tions, is that the high Keewatin and Labrador domes of the smaller melt-producing area of the Laurentide Ice Sheet. The Laurentide Ice Sheet and ice advance across the Great Lakes drying hypothesis has received further critiques from Teller basins generated a new transcontinental drainage divide that (2013) and Breckenridge(2015) based on field evidence, added regions that currently flow into Hudson Bay and the with strandline mapping by Breckenridge(2015) indicating Saint Lawrence River to the LGM Mississippi drainage basin that early eastward outflow from Lake Agassiz is plausible, (Fig.7). The patterns of post-LGM drainage basin evolution, and that the Moorhead Phase rapid lake-level drop (see Lep- as well as regional-scale changes in drainage basin extents, per et al., 2013) and associated meltwater outflow occurred vary more significantly among model-based reconstructions. during northwesterly flow rerouting to the Mackenzie River. Identifying differences between the model-driven and data- Regardless of where Lake Agassiz flowed, or whether it be- driven drainage basin reconstructions can guide efforts to came a closed basin (the flow routing presented here does not produce a more accurate depiction of the last deglaciation. allow closed basins to exist: see Sect. 2.3.2), isotopic data At the LGM, ICE-5G and G12 have the most severe misfits from the Gulf of Mexico show unequivocally that meltwater with data. ICE-5G sends a large fraction of what data indi- discharge to the Mississippi dropped precipitously ∼ 12.9 ka cate should be Mississippi-River-bound drainage to the Hud- (Williams et al., 2012; Wickert et al., 2013). This means that son River, an artifact fixed in ICE-6G with its more realistic the ICE-5G, ANU, and G12 reconstructions miss the tim- ice-dome structure. G12 at the LGM has Mackenzie River ing of the most major drainage rearrangement in the North drainage routed towards the Yukon River, though time steps American deglaciation. In G12, this is due in part to the too shortly before and after the LGM have the proper drainage early retreat of the southeastern Laurentide margin. Over-

Earth Surf. Dynam., 4, 831–869, 2016 www.earth-surf-dynam.net/4/831/2016/ A. D. Wickert: North American deglacial water routing 857 all, ICE-3G and ICE-6G perform the best at 12.5 ka, with ern Mackenzie River drainage basin. Flow from the eastern the former sending Lake Agassiz meltwater to the Gulf of modern Mackenzie River drainage basin did not return to the Saint Lawrence and the latter sending Lake Agassiz melt- Mackenzie River until significant isostatic rebound tilted the water to the Mackenzie River and Beaufort Sea; in both land surface to the west. New cores from the Beaufort Sea ICE-3G and ICE-6G, the Mackenzie and Saint Lawrence record increased δ18O between 11 and 9 ka when corrected received proximal meltwater inputs. Either of these rout- for ice-sheet volume (Keigwin and Driscoll, 2014; L. Keig- ings at 12.5 ka is within error of the 12.18 ± 0.48 ka date for win, personal communication, 2016). This increase cannot drainage rerouting from the Saint Lawrence to the Mackenzie be reconciled with changing climate but would be consis- (Breckenridge, 2015) and is also consistent within error of tent with a decrease in drainage basin area. ICE-5G, ICE- +0.3 the 12.3−0.4 ka early peak in radiocarbon dates on roots from 6G, G12, and the ANU model each include this reduced the Moorhead Phase low in the Lake Agassiz basin (Fisher drainage basin area at different times (see videos in the Sup- et al., 2008). plement). Therefore, the temporary reduction in Mackenzie Ice retreat during the terminal Pleistocene and early River drainage area is sensitive to ice-sheet geometry and/or Holocene reveals excessive isostatic adjustment in some of solid-Earth rheology, making it a useful constraint on ice- the models. The overly large isostatic depression produced sheet reconstructions. Among the models studied here, G12 by G12/VM2 causes the Great Lakes system to become a fits this timing the best by having a greatly reduced Holocene marine embayment following ice retreat, which leads to an drainage basin area from 11.0 to 8.2 ka, although it also erroneous 9.6 ka rerouting of Saint Lawrence drainage to produced a small Mackenzie River drainage basin at earlier Hudson Strait. Excess isostatic subsidence produced by ICE- times. 5G/VM2, G12/VM2, and ICE-6G/VM5a also results in a north-flowing upper Missouri River. While there is a plausi- 5.2 Future directions: new ice-sheet reconstructions ble course for northerly flow from the upper Missouri River and coupling with hydrologic, climate, isotopic, into the preglacial Heart River at Bismarck, North Dakota archaeological, and sediment transport models and (Todd, 1914; Leonard, 1916; Bluemle, 2006), at present no observations field studies indicate drainage rearrangement following the LGM. The work presented here is a necessary starting point The early Holocene in northern North America was char- for future studies that investigate the interactions between acterized by retreat of the Laurentide Ice Sheet that led to deglacial surface-water hydrology and other components expansion of proglacial lakes, setting the stage for the short- of the integrated Earth system. Combining these meltwa- lived and rapid 8.2 ka cooling event when the ice dam sep- ter routing patterns with proglacial lakes and changing arating Lake Agassiz–Ojibway and the sea collapsed and land cover will allow equations designed to predict sed- the lake rapidly drained into the ocean (Barber et al., 1999; iment yield from large catchments to be employed for Breckenridge et al., 2012; Gregoire et al., 2012). None of the a continental-scale paleo-sediment-discharge reconstruction models perform particularly well at this time, though some (following Overeem et al., 2005; Kettner and Syvitski, 2008; observations hold promise. ICE-3G has altogether too much Pelletier, 2012; Cohen et al., 2013), with modern control pro- ice retreat. ICE-5G, ICE-6G, and ANU have a drainage di- vided by present-day sediment yields (Table1). These sedi- vide that splits the east–west Saint Lawrence drainage basin, ment discharges can be compared with records of deposition which at this time extended west to the Canadian Rockies (e.g., Andrews and Dunhill, 2004; Breckenridge, 2007; Rit- (Teller and Leverington, 2004; Breckenridge et al., 2012; tenour et al., 2007; Williams et al., 2010) and geomorphic Stroup et al., 2013). G12 has this continuous drainage basin, change (e.g., Dury, 1964; Reusser et al., 2006; Knox, 2007; but instead of flowing into the Saint Lawrence River, water Bettis et al., 2008; Anderson, 2015). The need to properly in its drainage reconstruction flows to the Arctic Ocean via compute past lake and land cover motivates continued work a gap at the Manitoba–Saskatchewan border along the south- with climate- and water-balance models (e.g., Collins et al., western corner of the retreating ice sheet (Fig. 10). While this 2006; Matsubara and Howard, 2009; Liu et al., 2009; He, could be a route for the ∼ 8.6 ka flood associated with par- 2011; Blois et al., 2013; Fan et al., 2013; Ivanovic´ et al., tial drainage of Lake Agassiz–Ojibway (Breckenridge et al., 2016a), which can in turn be used to improve past drainage 2012, p. 53), it is not the long-term drainage route for these basin and discharge reconstructions. These, together with pa- lakes through the Kinojévis spillway system and into the Ot- leogeographic reconstructions such as those presented here, tawa River (Vincent and Hardy, 1977; Veillette, 1994). can be used to reconstruct areas of archaeological inter- A tiny early Holocene Mackenzie River basin appears in est, either as changes in shoreline positions and topogra- ICE-5G, ICE-6G, and G12, and the ANU model also pro- phy (Clark et al., 2014) or as wholesale landscape recon- duces a small Mackenzie drainage basin at 8.75 and 8.45 ka. structions that incorporate site-potential modeling (Fedje and In these models, rapid retreat of the western margin of the Christensen, 1999; Mandryk et al., 2001; Monteleone, 2013; Laurentide Ice Sheet exposed an eastward-tilting isostatically Monteleone et al., 2013; Dixon and Monteleone, 2014). On depressed land surface in the eastern portion of the mod- a global scale, several current flow-routing algorithms could www.earth-surf-dynam.net/4/831/2016/ Earth Surf. Dynam., 4, 831–869, 2016 858 A. D. Wickert: North American deglacial water routing be made global for better integration with ice-sheet, climate, study were not calibrated to drainage basins or meltwater and GIA models (Metz et al., 2011; Qin and Zhan, 2012; pathways, which leave behind geomorphic and stratigraphic Braun and Willett, 2013; Huang and Lee, 2013; Schwang- markers that can be used to characterize the plausibility hart and Scherler, 2014), with the possibility to include high- of hypothesized ice-thickness distributions and deglaciation resolution flow routing as part of a transient coupled GCM histories. instead of an a posteriori analysis, as is presented here. Fi- Significant scatter in the data–model fit among the ice- nally, high-resolution drainage routing schemes can connect sheet reconstructions tested leaves no clear “winner” among models of past climate, ice sheets, and drainage routing to these models, though a few concluding remarks can be made. oxygen isotopes in sediment cores (Wickert et al., 2013) ICE-6G is based on GIA and ice margins, but it also pos- The increasingly complete collection of such records from sesses an ice-dome structure that is consistent with geo- the North American continent and continental margin (e.g., logical evidence of ice-flow paths. Its generally good per- Hooke and Clausen, 1982; Remenda et al., 1994; Andrews formance is likely tied to this combined geological and et al., 1994; de Vernal et al., 1996; Birks et al., 2007; Carlson geophysical basis. G12, while only loosely calibrated to et al., 2007a, 2009; Breckenridge and Johnson, 2009; Lopes match the outlines of the North American ice-sheet complex, and Mix, 2009; Obbink et al., 2010; Brown, 2011; Hoffman nonetheless reproduces many changes in continental-scale et al., 2012; Williams et al., 2012; Gibb et al., 2014; Tay- drainage. Its basis in a climate-driven process-based numeri- lor et al., 2014; Ferguson and Jasechko, 2015; Hladyniuk cal model helped it to generate periods of high and low river and Longstaffe, 2016) is opening new possibilities in isotopic discharge that were largely in agreement with the geological studies of whole-ice-sheet mass balance. record, and permitted tightly spaced output time steps that Pursuit of these targets does not preclude the search for a enable better comparison between the reconstruction and the more representative ice-sheet reconstruction and better ways geologic record. Unexpectedly, the venerable ICE-3G per- to integrate models and data. New reduced-complexity mod- formed the best in the Kolmogorov–Smirnov test for times eling rooted in ice-margin mapping and ice physics (Gowan of high and low discharge; this was due largely to its strong et al., 2016a,b) spans the gap between the empirical and fit with the data-driven Mackenzie River record. modeling approaches highlighted here. Surface energy bal- Beyond the focus on deglaciation and ice-sheet geome- ance calculations (Ullman et al., 2015) also elucidate which try, the dynamic deglacial drainage basins of North America part of the ice sheet was melting, and when, providing a are significant in their own right. The new paleohydrographs testable link to records of glacial meltwater input. Contin- presented here provide new insights into the ice-age legacy ued development of numerical ice-sheet models (e.g., Corn- imprinted on the major rivers of the continent and quantify ford et al., 2013; Le Morzadec et al., 2015) provides new av- the mid-20th-century observation that many North American enues to reconstruct past ice sheets with climate inputs from rivers are “underfit” in their glacial-meltwater-carved valleys GCM ensembles (Gregoire et al., 2016) and route meltwa- (Dury, 1964). These paleohydrographs and drainage basin ter through an evolving deglacial drainage network (Tarasov changes provide the paleogeographic framework for new ad- and Peltier, 2006; Tarasov et al., 2012; Ivanovic´ et al., 2014, vances in paleohydrology and fluvial geomorphology. 2016b). Such fully coupled models, if run at a high enough The goal of this work is to move towards a self-consistent spatial and temporal resolution, have the potential to recre- paleogeographic framework within which models and geo- ate short-term flooding events that have left a strong mark in logic records may be quantitatively compared to build new the geologic record of deglaciation (e.g., Kehew and Teller, insights into past glacial systems and climate–ice-sheet in- 1994; Breckenridge, 2007; Murton et al., 2010). Climate and teractions. When integrated more broadly into studies of ice-sheet models may also be isotope-enabled for compari- North American deglaciation, these time-varying deglacial son with isotopic data from geological archives (Caley et al., drainage basins and paleohydrographs are poised to improve 2014; Ferguson and Jasechko, 2015; Gasson et al., 2016). reconstructions of past climate and ice-sheet geometry. Each of these approaches presents new opportunities to test deglacial drainage patterns and converge towards a solution for the shape and evolution of the North American ice-sheet 7 Code and model output availability complex that integrates data from glacial, terrestrial, and marine systems. All computer code for this research is available on A. Wickert’s GitHub repository, at https://github.com/ awickert/ice-age-rivers, and a fork of this repository is 6 Conclusions available from the University of Minnesota Earth Sur- face Processes GitHub organization at https://github.com/ Most ice-sheet reconstructions, including all those examined umn-earth-surface/ice-age-rivers. The drainage basin and here, are built to fit sea-level records, glacial isostatic ad- discharge histories developed for this work are provided justment measurements, and/or moraine positions; some are there as well, and are deposited for long-term storage at the built using ice-physics-based simulations. The models in this Data Repository for the University of Minnesota (DRUM),

Earth Surf. Dynam., 4, 831–869, 2016 www.earth-surf-dynam.net/4/831/2016/ A. D. Wickert: North American deglacial water routing 859 at http://doi.org/10.13020/D6D01H (Wickert, 2016). Com- Anders, M. D., Pederson, J. L., Rittenour, T. M., Sharp, W. D., plete time series of maps of deglacial drainage such Gosse, J. C., Karlstrom, K. E., Crossey, L. J., Goble, R. J., as those shown in Figs.7–10 are also available from Stockli, L., and Yang, G.: Pleistocene geomorphology and DRUM. These images portray past elevations and sea geochronology of eastern Grand Canyon: linkages of landscape level (background map), drainage basin extents (black out- components during climate changes, Quaternary Sci. Rev., 24, lines), ice extents (white-filled contour), and rivers with 2428–2448, doi:10.1016/j.quascirev.2005.03.015, 2005. Anderson, J.: History of the Missouri River Valley from the Late discharge that is ≥ 1000 m3 s−1 (blue lines). Movies cre- Pleistocene to Present: Climatic vs . Tectonic Forcing on Valley ated from these images are stored at DRUM and are also Architecture, MS Thesis, Texas Christian University, 2015. available as a supplement to this article. The compiled Anderson, L. S., Roe, G. H., and Anderson, R. S.: The effects of mean global precipitation and evapotranspiration data prod- interannual climate variability on the moraine record, Geology, ucts may be downloaded from https://github.com/awickert/ 42, 55–58, doi:10.1130/G34791.1, 2014. global-precipitation-evapotranspiration. Software is avail- Anderson, R. C.: Reconstruction of preglacial drainage and its di- able under the GNU General Public License v3, and climate version by earliest glacial forebulge in the upper Mississippi Val- data compilations, model outputs, and all other non-software ley region, Geology, 16, 254–257, 1988. products are licensed as Creative Commons Attribution- Andersson, A., Fennig, K., Klepp, C., Bakan, S., Graßl, H., and ShareAlike (CC-BY-SA 3.0). Schulz, J.: The Hamburg Ocean Atmosphere Parameters and Fluxes from Satellite Data – HOAPS-3, Earth Syst. Sci. Data, 2, 215–234, doi:10.5194/essd-2-215-2010, 2010. The Supplement related to this article is available online Andersson, A., Klepp, C., Fennig, K., Bakan, S., Grassl, H., and at doi:10.5194/esurf-4-831-2016-supplement. Schulz, J.: Evaluation of HOAPS-3 Ocean Surface Freshwa- ter Flux Components, J. Appl. Meteorol. Clim., 50, 379–398, doi:10.1175/2010JAMC2341.1, 2011. Andrews, J. T. and Dunhill, G.: Early to mid-Holocene At- lantic water influx and deglacial meltwater events, Beau- Acknowledgements. Jerry Mitrovica provided GIA model out- fort Sea slope, Arctic Ocean, Quaternary Res., 61, 14–21, puts and Feng He supplied TraCE-21K GCM outputs, both of which doi:10.1016/j.yqres.2003.08.003, 2004. were crucial to this work. CMAP precipitation data were provided Andrews, J. T. and MacLean, B.: Hudson Strait ice by the NOAA/OAR/ESRL PSD, Boulder, Colorado, USA, from streams: a review of stratigraphy, chronology and links their web site at http://www.esrl.noaa.gov/psd/. Ruža Ivanovic,´ with North Atlantic Heinrich events, Boreas, 32, 4–17, Lauren Gregoire, and Kelly Monteleone helped with stimulating doi:10.1080/03009480310001010, 2003. discussion and ideas about where to take this work in the future. Andrews, J. T. and Tedesco, K.: Detrital carbonate-rich sed- Conversations with Bob Anderson about glacial impacts on the iments, northwestern Labrador Sea: Implications for ice- morphology of large rivers inspired this work, and reviews from sheet dynamics and iceberg rafting (Heinrich) events in the Andy Breckenridge and Lev Tarasov as well as comments from North Atlantic, Geology, 20, 1087–1090, doi:10.1130/0091- Anders Carlson improved the final publication. A. D. Wickert was 7613(1992)020<1087:DCRSNL>2.3.CO;2, 1992. supported by the US Department of Defense through the National Andrews, J. T., Erlenkeuser, H., Tedesco, K., Aksu, A. E., and Jull, Defense Science & Engineering Graduate Fellowship Program, the A.: Late Quaternary (Stage 2 and 3) Meltwater and Heinrich US National Science Foundation Graduate Research Fellowship Events, Northwest Labrador Sea, Quaternary Res., 41, 26–34, under grant no. DGE 1144083, the Emmy Noether Programme doi:10.1006/qres.1994.1003, 1994. of the Deutsche Forschungsgemeinschaft (DFG) through funds Andrews, J. T., Keigwin, L., Hall, F., and Jennings, A. E.: Abrupt awarded to T. Schildgen under grant no. SCHI 1241/1-1, and deglaciation events and Holocene palaeoceanography from high- start-up funds from the University of Minnesota. resolution cores, Cartwright Saddle, Labrador Shelf, Canada, J. Quaternary Sci., 14, 383–397, doi:10.1002/(SICI)1099- Edited by: J. K. Hillier 1417(199908)14:5<383::AID-JQS464>3.0.CO;2-J, 1999. Reviewed by: A. Breckenridge and L. Tarasov Antevs, E.: The recession of the last ice sheet in New England, American Geographical Society of New York: The Conde Nast Press, Greenwich, CT, USA, 1922. References Argus, D. F. and Peltier, W. R.: Constraining models of postglacial rebound using space geodesy: a detailed assessment of model Adler, R. F., Huffman, G. J., Chang, A., Ferraro, R., Xie, ICE-5G (VM2) and its relatives, Geophys. J. Int., 181, 697–723, P.-P., Janowiak, J., Rudolf, B., Schneider, U., Curtis, 2010. S., Bolvin, D., Gruber, A., Susskind, J., Arkin, P., and Argus, D. F., Peltier, W. R., Drummond, R., and Moore, A. W.: The Nelkin, E.: The Version-2 Global Precipitation Climatology Antarctica component of postglacial rebound model ICE-6G_C Project (GPCP) Monthly Precipitation Analysis (1979– (VM5a) based on GPS positioning, exposure age dating of ice Present), J. Hydrometeorol., 4, 1147–1167, doi:10.1175/1525- thicknesses, and relative sea level histories, Geophys. J. Int., 198, 7541(2003)004<1147:TVGPCP>2.0.CO;2, 2003. 537–563, doi:10.1093/gji/ggu140, 2014. Amante, C. and Eakins, B. W.: ETOPO1 1 Arc-Minute Global Asmerom, Y., Polyak, V. J., and Burns, S. J.: Variable winter mois- Relief Model: Procedures, Data Sources and Analysis, NOAA ture in the southwestern United States linked to rapid glacial Technical Memorandum NESDIS NGDC-24, 2009. www.earth-surf-dynam.net/4/831/2016/ Earth Surf. Dynam., 4, 831–869, 2016 860 A. D. Wickert: North American deglacial water routing

climate shifts, Nat. Geosci., 3, 114–117, doi:10.1038/ngeo754, vial incision and landscape evolution, Geomorphology, 180–181, 2010. 170–179, doi:10.1016/j.geomorph.2012.10.008, 2013. Atwater, B. F.: Periodic floods from glacial Lake Missoula Breckenridge, A.: The Lake Superior varve stratigraphy and impli- into the Sanpoil arm of glacial Lake Columbia, northeast- cations for eastern Lake Agassiz outflow from 10,700 to 8900 ern Washington, Geology, 12, 464–467, doi:10.1130/0091- cal ybp (9.5–8.0 {14}C ka), Palaeogeogr. Palaeocl., 246, 45–61, 7613(1984)12<464:PFFGLM>2.0.CO;2, 1984. doi:10.1016/j.palaeo.2006.10.026, 2007. Austermann, J., Mitrovica, J. X., Latychev, K., and Milne, G. A.: Breckenridge, A.: The Tintah-Campbell gap and implications Barbados-based estimate of ice volume at Last Glacial Max- for glacial Lake Agassiz drainage during the Younger imum affected by subducted plate, Nat. Geosci., 6, 553–557, Dryas cold interval, Quaternary Sci. Rev., 117, 124–134, doi:10.1038/ngeo1859, 2013. doi:10.1016/j.quascirev.2015.04.009, 2015. Barber, D. C., Jennings, A. E., Andrews, J. T., Kerwin, M. W., and Breckenridge, A. and Johnson, T. C.: Paleohydrology of the upper Morehead, M. D.: Forcing of the cold event of 8,200 years ago Laurentian Great Lakes from the late glacial to early Holocene, by catastrophic drainage of Laurentide lakes, Nature, 400, 13–15, Quaternary Res., 71, 397–408, doi:10.1016/j.yqres.2009.01.003, doi:10.1038/22504, 1999. 2009. Bell, R.: On glacial phenomena in Canada, Bull. Geol. Soc. Am., 1, Breckenridge, A., Lowell, T. V., Stroup, J. S., and Evans, G.: A 287–310, 1889. review and analysis of varve thickness records from glacial Lake Benito, G. and O’Connor, J. E.: Number and size of last- Ojibway (Ontario and Quebec, Canada), Quatern. Int., 260, 43– glacial Missoula floods in the Columbia River valley be- 54, doi:10.1016/j.quaint.2011.09.031, 2012. tween the Pasco Basin, Washington, and Portland, Ore- Bretz, J. H.: The channeled scablands of the Columbia Plateau, J. gon, Geol. Soc. Am. Bull., 115, 624–638, doi:10.1130/0016- Geol., 31, 617–649, 1923. 7606(2003)115<0624:NASOLM>2.0.CO;2, 2003. Bretz, J. H.: The Lake Missoula floods and the channeled scabland, Benson, L., Burdett, J., Lund, S., Kashgarian, M., and Mensing, S.: J. Geol., 77, 505–543, 1969. Nearly synchronous climate change in the Northern Hemisphere British Oceanographic Data Centre: The GEBCO_08 Grid, version during the last glacial termination, Nature, 388, 263–265, 1997. 20100927, 2010. Bentley, S.: Fluvial Sediment Delivery to Hudson Bay in a Chang- Broecker, W. S. and Farrand, W. R.: Radiocarbon age of the two ing Climate: Projections and Hypotheses, in: ArcticNet Annual creeks forest bed, Wisconsin, Bull. Geol. Soc. Am., 74, 795– Scientific Meeting, 12–15 December, Victoria, British Columbia, 802, doi:10.1130/0016-7606(1963)74[795:RAOTTC]2.0.CO;2, Canada, 2006. 1963. Bettis, E. A., Benn, D. W., and Hajic, E. R.: Landscape evolution, Broecker, W. S., Kennett, J. P., Flower, B. P., Teller, J. T., Trum- alluvial architecture, environmental history, and the archaeolog- bore, S., Bonani, G., and Wolfli, W.: Routing of meltwater from ical record of the Upper Mississippi River Valley, Geomorphol- the Laurentide Ice Sheet during the Younger Dryas cold episode, ogy, 101, 362–377, doi:10.1016/j.geomorph.2008.05.030, 2008. Nature, 341, 318–321, doi:10.1038/341318a0, 1989. Birks, S. J., Edwards, T. W. D., and Remenda, V. H.: Isotopic evo- Bromwich, D. H., Toracinta, E. R., Wei, H., Oglesby, R. J., Fas- lution of Glacial Lake Agassiz: New insights from cellulose and took, J. L., Hughes, T. J., Group, P. M., Polar, B., Program, porewater isotopic archives, Palaeogeogr. Palaeocl., 246, 8–22, A. S., Springs, C., Studies, C., Bromwich, D. H., Toracinta, doi:10.1016/j.palaeo.2006.10.024, 2007. E. R., Wei, H., Oglesby, R. J., Fastook, J. L., Hughes, T. J., Blois, J. L., Williams, J. W., Fitzpatrick, M. C., Ferrier, S., Veloz, Group, P. M., Polar, B., Program, A. S., Springs, C., Stud- S. D., He, F., Liu, Z., Manion, G., and Otto-Bliesner, B.: Model- ies, C., Bromwich, D. H., Toracinta, E. R., Wei, H., Oglesby, ing the climatic drivers of spatial patterns in vegetation composi- R. J., Fastook, J. L., and Hughes, T. J.: Polar MM5 Sim- tion since the Last Glacial Maximum, Ecography, 36, 460–473, ulations of the Winter Climate of the Laurentide Ice Sheet doi:10.1111/j.1600-0587.2012.07852.x, 2013. at the LGM, J. Climate, 17, 3415–3433, doi:10.1175/1520- Bluemle, J. P.: Pleistocene Drainage Development in North 0442(2004)017<3415:PMSOTW>2.0.CO;2, 2004. Dakota, Geol. Soc. Am. Bull., 83, 2189, doi:10.1130/0016- Brown, E. A.: Initial Ablation of the Laurentide Ice Sheet Based 7606(1972)83[2189:PDDIND]2.0.CO;2, 1972. on Gulf of Mexico Sediments, MS thesis, University of South Bluemle, J. P.: Origin of the Missouri River in North Dakota, in: Florida, St. Petersburg, FL, USA, 2011. Quaternary geology of the Missouri River Valley and adjacent Brugger, K. A.: Climate in the Southern Sawatch Range and Elk areas in northwest-central North Dakota, edited by: Manz, L. A., Mountains, Colorado, U.S.A., during the Last Glacial Maximum: Midwest Friends of the Pleistocene, North Dakota Geological Inferences Using a Simple Degree-Day Model, Arct. Antarct. Survey, 2006. Alp. Res., 42, 164–178, doi:10.1657/1938-4246-42.2.164, 2010. Blum, M. D., Guccione, M. J., Wysocki, D. A., Robnett, Caley, T., Roche, D. M., Waelbroeck, C., and Michel, E.: Oxy- P. C., and Rutledge, E. M.: Late Pleistocene evolution of the gen stable isotopes during the Last Glacial Maximum climate: lower Mississippi River valley, southern Missouri to Arkansas, perspectives from data-model (iLOVECLIM) comparison, Clim. Geol. Soc. Am. Bull., 112, 221–235, doi:10.1130/0016- Past, 10, 1939–1955, doi:10.5194/cp-10-1939-2014, 2014. 7606(2000)112<221:LPEOTL>2.0.CO;2, 2000. Carlson, A. E.: Geochemical constraints on the Laurentide Ice Sheet Blumentritt, D. J., Wright, H. E., and Stefanova, V.: Formation and contribution to Meltwater Pulse 1A, Quaternary Sci. Rev., 28, early history of Lakes Pepin and St. Croix of the upper Missis- 1625–1630, 2009. sippi River, J. Paleolimnol., 41, 545–562, 2009. Carlson, A. E. and Clark, P. U.: Comment: Radiocarbon deglacia- Braun, J. and Willett, S. D.: A very efficient O(n), implicit and par- tion chronology of the Thunder Bay, Ontario area and impli- allel method to solve the stream power equation governing flu-

Earth Surf. Dynam., 4, 831–869, 2016 www.earth-surf-dynam.net/4/831/2016/ A. D. Wickert: North American deglacial water routing 861

cations for ice sheet retreat patterns, Quaternary Sci. Rev., 28, Cohen, M. J., Henges-Jeck, C., and Castillo-Moreno, G.: A prelim- 2546–2547, doi:10.1016/j.quascirev.2009.02.025, 2009. inary water balance for the Colorado River delta, 1992–1998, J. Carlson, A. E. and Clark, P. U.: Ice sheet sources of sea level rise Arid Environ., 49, 35–48, doi:10.1006/jare.2001.0834, 2001. and freshwater discharge during the last deglaciation, Rev. Geo- Cohen, S., Kettner, A. J., Syvitski, J. P. M., and Fekete, B. M.: phys., 50, RG4007, doi:10.1029/2011RG000371, 2012. WBMsed, a distributed global-scale riverine sediment flux Carlson, A. E., Clark, P. U., Haley, B. A., Klinkhammer, G. P., Sim- model: Model description and validation, Comput. & Geosci., mons, K., Brook, E. J., and Meissner, K. J.: Geochemical prox- 53, 80–93, doi:10.1016/j.cageo.2011.08.011, 2013. ies of North American freshwater routing during the Younger COHMAP members: Climatic Changes of the Last 18,000 Years: Dryas cold event, P. Natl. Acad. Sci. USA, 104, 6556–6561, Observations and Model Simulations, Science, 241, 1043–1052, doi:10.1073/pnas.0611313104, 2007a. doi:10.1126/science.241.4869.1043, 1988. Carlson, A. E., Clark, P. U., Raisbeck, G. M., and Brook, Collins, W. D., Bitz, C. M., Blackmon, M. L., Bonan, G. B., E. J.: Rapid holocene deglaciation of the labrador sector Bretherton, C. S., Carton, J. A., Chang, P., Doney, S. C., Hack, of the Laurentide Ice Sheet, J. Climate, 20, 5126–5133, J. J., Henderson, T. B., Kiehl, J. T., Large, W. G., McKenna, doi:10.1175/JCLI4273.1, 2007b. D. S., Santer, B. D., and Smith, R. D.: The Community Climate Carlson, A. E., LeGrande, A. N., Oppo, D. W., Came, R. E., System Model Version 3 (CCSM3), J. Climate, 19, 2122–2143, Schmidt, G. A., Anslow, F. S., Licciardi, J. M., and Obbink, doi:10.1175/JCLI3761.1, 2006. E. A.: Rapid early Holocene deglaciation of the Laurentide ice Condron, A. and Winsor, P.: Meltwater routing and the sheet, Nat. Geosci., 1, 620–624, doi:10.1038/ngeo285, 2008. Younger Dryas, P. Natl. Acad. Sci. USA, 109, 19928–19933, Carlson, A. E., Clark, P. U., Haley, B. A., and Klinkham- doi:10.1073/pnas.1207381109, 2012. mer, G. P.: Routing of western Canadian Plains runoff dur- Cook, K. L., Whipple, K. X., Heimsath, A. M., and Hanks, T. C.: ing the 8.2 ka cold event, Geophys. Res. Lett., 36, 1–5, Rapid incision of the Colorado River in Glen Canyon - in- doi:10.1029/2009GL038778, 2009. sights from channel profiles, local incision rates, and modeling Carriquiry, J. D. and Sánchez, A.: Sedimentation in the Colorado of lithologic controls, Earth Surf. Proc. Land., 1010, 994–1010, River delta and Upper Gulf of California after nearly a century doi:10.1002/esp.1790, 2009. of discharge loss, Mar. Geol., 158, 125–145, doi:10.1016/S0025- Cornford, S. L., Martin, D. F., Graves, D. T., Ranken, D. F., 3227(98)00189-3, 1999. Le Brocq, A. M., Gladstone, R. M., Payne, A. J., Ng, E. G., Carson, M. A., Jasper, J. N., and Conly, F. M.: Magnitude and and Lipscomb, W. H.: Adaptive mesh, finite volume model- sources of sediment input to the Mackenzie Delta, Northwest ing of marine ice sheets, J. Comput. Phys., 232, 529–549, Territories, 1974–94, Arctic, 51, 116–124, 1998. doi:10.1016/j.jcp.2012.08.037, 2013. Catto, N., Liverman, D. G. E., Bobrowsky, P. T., and Rutter, N.: Cronin, T. M., Rayburn, J., Guilbault, J.-P., Thunell, R., and Franzi, Laurentide, cordilleran, and montane glaciation in the west- D.: Stable isotope evidence for glacial lake drainage through the ern Peace River – Grande Prairie region, Alberta and British St. Lawrence Estuary, eastern Canada, ∼ 3.1–12.9 ka, Quatern. Columbia, Canada, Quatern. Int., 32, 21–32, doi:10.1016/1040- Int., 260, 55–65, doi:10.1016/j.quaint.2011.08.041, 2012. 6182(95)00061-5, 1996. Cuffey, K. M. and Paterson, W. S. B.: The physics of glaciers, Aca- Chen, M., Xie, P., Janowiak, J. E., and Arkin, P. A.: Global Land demic Press, Oxford, UK, 4th edn., 2010. Precipitation: A 50-yr Monthly Analysis Based on Gauge Ob- Currey, D. R.: Quaternary palaeolakes in the evolution of semidesert servations, J. Hydrometeorol., 3, 249–266, doi:10.1175/1525- basins, with special emphasis on Lake Bonneville and the 7541(2002)003<0249:GLPAYM>2.0.CO;2, 2002. Great Basin, U.S.A, Palaeogeogr. Palaeoclimatol. Palaeoecol., Clague, J. J. and James, T. S.: History and isostatic effects of the last 76, 189–214, doi:10.1016/0031-0182(90)90113-L, 1990. ice sheet in southern British Columbia, Quaternary Sci. Rev., 21, Curry, B. B.: Evidence at Lomax, Illinois, for Mid-Wisconsin 71–87, doi:10.1016/S0277-3791(01)00070-1, 2002. (40,000 yr BP) Position of the Des Moines Lobe and for Diver- Clark, J., Mitrovica, J. X., and Alder, J.: Coastal paleogeography sion of the Mississippi River by the Lake Michigan Lobe (20,350 of the California-Oregon-Washington and Bering Sea continen- yr BP)∗ 1, Quaternary Res., 50, 128–138, 1998. tal shelves during the latest Pleistocene and Holocene: Implica- Curry, B. B., Hajic, E. R., Clark, J. A., Befus, K. M., tions for the archaeological record, J. Archaeol. Sci., 52, 12–23, Carrell, J. E., and Brown, S. E.: The Kankakee Torrent doi:10.1016/j.jas.2014.07.030, 2014. and other large meltwater flooding events during the last Clark, P. U., Marshall, S. J., Clarke, G. K. C., Hostetler, S. W., Lic- deglaciation, Illinois, USA, Quaternary Sci. Rev., 90, 22–36, ciardi, J. M., and Teller, J. T.: Freshwater forcing of abrupt cli- doi:10.1016/j.quascirev.2014.02.006, 2014. mate change during the last glaciation, Science, 293, 283–287, de Vernal, A., Hillaire-Marcel, C., and Bilodeau, G.: Reduced melt- doi:10.1126/science.1062517, 2001. water outflow from the Laurentide ice margin during the Younger Clark, P. U., Dyke, A. S., Shakun, J. D., Carlson, A. E., Clark, Dryas, Nature, 381, 774–777, doi:10.1038/381774a0, 1996. J., Wohlfarth, B., Mitrovica, J. X., Hostetler, S. W., and Mc- Deschamps, P., Durand, N., Bard, E., Hamelin, B., Camoin, Cabe, A. M.: The Last Glacial Maximum, Science, 325, 710– G., Thomas, A. L., Henderson, G. M., Okuno, J., and 714, doi:10.1126/science.1172873, 2009. Yokoyama, Y.: Ice-sheet collapse and sea-level rise at the Clayton, L. and Moran, S. R.: Chronology of late wisconsinan Bølling warming 14,600 years ago, Nature, 483, 559–564, glaciation in middle North America, Quaternary Sci. Rev., 1, 55– doi:10.1038/nature10902, 2012. 82, doi:10.1016/0277-3791(82)90019-1, 1982. Dierckx, P.: Curve and surface fitting with splines, Oxford Univer- sity Press, 1993.

www.earth-surf-dynam.net/4/831/2016/ Earth Surf. Dynam., 4, 831–869, 2016 862 A. D. Wickert: North American deglacial water routing

Dixon, J. E. and Monteleone, K.: Gateway to the Americas: Under- Fisher, T. G.: Chronology of glacial Lake Agassiz meltwater routed water Archeological Survey in Beringia and the North Pacific, in: to the Gulf of Mexico, Quaternary Res., 59, 271–276, 2003. Prehistoric Archaeology on the Continental Shelf, chap. 6, 95– Fisher, T. G.: Abandonment chronology of glacial Lake Agas- 114, Springer, New York, doi:10.1007/978-1-4614-9635-9_6, siz’s Northwestern outlet, Palaeogeogr. Palaeocl., 246, 31–44, 2014. doi:10.1016/j.palaeo.2006.10.031, 2007. Dolgopolova, E. N. and Isupova, M. V.: Water and sediment dynam- Fisher, T. G. and Souch, C.: Northwest outlet channels of Lake ics at Saint Lawrence River mouth, Water Resour., 38, 453–469, Agassiz, isostatic tilting and a migrating continental drainage doi:10.1134/S009780781104004X, 2011. divide, Saskatchewan, Canada, Geomorphology, 25, 57–73, Doyon, P.: Seasonal Structure of the Gulf of St. Lawrence Upper- doi:10.1016/S0169-555X(98)00028-2, 1998. Layer Thermohaline Fields during the Ice-free Months, MS The- Fisher, T. G., Yansa, C. H., Lowell, T. V., Lepper, K., Hajdas, I., sis, McGill University, 1996. and Ashworth, A.: The chronology, climate, and confusion of Dreimanis, A. and Goldthwait, R. P.: Wisconsin Glaciation in the the Moorhead Phase of glacial Lake Agassiz: new results from Huron, Erie, and Ontario Lobes, in: GSA Memoirs: The Wiscon- the Ojata Beach, North Dakota, USA, Quaternary Sci. Rev., 27, sinan Stage, 136, 71–106, Geological Society of America, Boul- 1124–1135, doi:10.1016/j.quascirev.2008.02.010, 2008. der, CO, USA, doi:10.1130/MEM136-p71, 1973. Flock, M. A.: The late Wisconsinan Savanna Terrace in tributaries Duk-Rodkin, A., Hughes, O. L., Survey, G., and Calgary, S. to the upper Mississippi River, Quaternary Res., 20, 165–176, N. W.: Tertiary-quaternary drainage of the Pre-glacial Macken- doi:10.1016/0033-5894(83)90075-3, 1983. zie basin, Quatern. Int., 22–23, 221–241, doi:10.1016/1040- Flower, B. P., Hastings, D. W., Hill, H. W., and Quinn, T. M.: Phas- 6182(94)90015-9, 1994. ing of deglacial warming and Laurentide Ice Sheet meltwater in Dury, G. H.: Subsurface Exploration and Chronology of under- the Gulf of Mexico, Geology, 32, 597, doi:10.1130/G20604.1, fit streams, Professional Paper 452-B, U.S. Geological Survey, 2004. Washington, DC, 1964. Galloway, W. E., Whiteaker, T. L., and Ganey-Curry, P.: History Dutton, A. and Lambeck, K.: Ice Volume and Sea Level of Cenozoic North American drainage basin evolution, sediment During the Last Interglacial, Science, 337, 216–219, yield, and accumulation in the Gulf of Mexico basin, Geosphere, doi:10.1126/science.1205749, 2012. 7, 938–973, doi:10.1130/GES00647.1, 2011. Dutton, A., Carlson, A. E., Long, A. J., Milne, G. A., Clark, P. U., Gasson, E., DeConto, R. M., Pollard, D., and Levy, R. H.: DeConto, R., Horton, B. P., Rahmstorf, S., and Raymo, M. E.: Dynamic Antarctic ice sheet during the early to mid- Sea-level rise due to polar ice-sheet mass loss during past warm Miocene, P. Natl. Acad. Sci. USA, 113, 3459–3464, periods, Science, 349, aaa4019, doi:10.1126/science.aaa4019, doi:10.1073/pnas.1516130113, 2016. 2015. Gates, L. D., Hagemann, S., and Golz, C.: Observed historical dis- Dyke, A. S.: An outline of North American deglaciation with charge data from major rivers for climate model validation, Tech. emphasis on central and northern Canada, in: Quaternary Rep. 307, Max-Planck-Institut für Meterologie, Hamburg, Ger- Glaciations–Extent and Chronology — Part II: North America, many, 2000. edited by: Ehlers, J. and Gibbard, P. L., vol. 2, Developments in Gibb, O. T., Hillaire-Marcel, C., and de Vernal, A.: Oceano- Quaternary Sciences, 373–424, 2004. graphic regimes in the northwest Labrador Sea since Ma- Dyke, A. S. and Prest, V. K.: Late Wisconsinan and Holocene His- rine Isotope Stage 3 based on dinocyst and stable iso- tory of the Laurentide Ice Sheet, Géographie physique et Quater- tope proxy records, Quaternary Sci. Rev., 92, 269–279, naire, 41, 237, doi:10.7202/032681ar, 1987. doi:10.1016/j.quascirev.2013.12.010, 2014. Dyke, A. S., Moore, A., and Robertson, L.: Deglaciation of North Gil, I. M., Keigwin, L. D., and Abrantes, F.: The deglacia- America, Open File 1574, Natural Resources Canada, Ottawa, tion over Laurentian Fan: History of diatoms, IRD, 2003. ice and fresh water, Quaternary Sci. Rev., 129, 57–67, Environment Canada: St. Lawrence River, available at: http://www. doi:10.1016/j.quascirev.2015.10.006, 2015. ec.gc.ca/stl/ (last access: 18 December 2015), 2013. Gilbert, G. K.: Lake Bonneville, Monograph 1, US Geological Sur- Evenson, E. B., Farrand, W. R., Eschman, D. F., Mickelson, D. M., vey, Washington, DC, 1890. and Maher, L. J.: Greatlakean Substage: A replacement for Godsey, H. S., Currey, D. R., and Chan, M. A.: New ev- Valderan Substage in the Lake Michigan basin, Quaternary Res., idence for an extended occupation of the Provo shoreline 6, 411–424, doi:10.1016/0033-5894(67)90005-1, 1976. and implications for regional climate change, Pleistocene Fairbanks, R. G.: A 17, 000-year glacio-eustatic sea level record: Lake Bonneville, Utah, USA, Quaternary Res., 63, 212–223, influence of glacial melting rates on the Younger Dryas event doi:10.1016/j.yqres.2005.01.002, 2005. and deep-ocean circulation, Nature, 342, 637–642, 1989. Godsey, H. S., Oviatt, C. G., Miller, D. M., and Chan, Fan, Y., Li, H., and Miguez-Macho, G.: Global Patterns M. A.: Stratigraphy and chronology of offshore to nearshore of Groundwater Table Depth, Science, 339, 940–943, deposits associated with the Provo shoreline, Pleistocene doi:10.1126/science.1229881, 2013. Lake Bonneville, Utah, Palaeogeogr. Palaeocl., 310, 442–450, Fedje, D. W. and Christensen, T.: Modeling Paleoshorelines and Lo- doi:10.1016/j.palaeo.2011.08.005, 2011. cating Early Holocene Coastal Sites in Haida Gwaii, Am. Antiq- Gomez, N., Gregoire, L. J., Mitrovica, J. X., and Payne, A. J.: uity, 64, 635, doi:10.2307/2694209, 1999. Laurentide-Cordilleran Ice Sheet saddle collapse as a contribu- Ferguson, G. and Jasechko, S.: The isotopic composition of the Lau- tion to meltwater pulse 1A, Geophys. Res. Lett., 42, 3954–3962, rentide Ice Sheet and fossil groundwater, Geophys. Res. Lett., 42, doi:10.1002/2015GL063960, 2015. 1–6, doi:10.1002/2015GL064106.Received, 2015.

Earth Surf. Dynam., 4, 831–869, 2016 www.earth-surf-dynam.net/4/831/2016/ A. D. Wickert: North American deglacial water routing 863

Gowan, E. J.: An assessment of the minimum timing of ice free He, F., Shakun, J. D., Clark, P. U., Carlson, A. E., Liu, Z., Otto- conditions of the western Laurentide Ice Sheet, Quaternary Sci. Bliesner, B. L., and Kutzbach, J. E.: Northern Hemisphere forc- Rev., 75, 100–113, doi:10.1016/j.quascirev.2013.06.001, 2013. ing of Southern Hemisphere climate during the last deglaciation, Gowan, E. J., Tregoning, P., Purcell, A., Lea, J., Fransner, O. J., Nature, 494, 81–85, doi:10.1038/nature11822, 2013. Noormets, R., and Dowdeswell, J. A.: ICESHEET 1.0: a program Hearty, P. J., Hollin, J. T., Neumann, A. C., O’Leary, M. J., and to produce paleo-ice sheet reconstructions with minimal assump- McCulloch, M.: Global sea-level fluctuations during the Last tions, Geosci. Model Dev., 9, 1673–1682, doi:10.5194/gmd-9- Interglaciation (MIS 5e), Quaternary Sci. Rev., 26, 2090–2112, 1673-2016, 2016a. doi:10.1016/j.quascirev.2007.06.019, 2007. Gowan, E. J., Tregoning, P., Purcell, A., Montillet, J.-P., and Mc- Heinrich, H.: Origin and consequences of cyclic ice rafting in the Clusky, S.: A model of the western Laurentide Ice Sheet, us- Northeast Atlantic Ocean during the past 130,000 years, Qua- ing observations of glacial isostatic adjustment, Quaternary Sci. ternary Res., 29, 142–152, doi:10.1016/0033-5894(88)90057-9, Rev., 139, 1–16, doi:10.1016/j.quascirev.2016.03.003, 2016b. 1988. GRASS Development Team: Geographic Resources Analysis Sup- Hemming, S. R.: Heinrich events: Massive late Pleistocene detritus port System (GRASS GIS) Software, 2015. layers of the North Atlantic and their global climate imprint, Rev. Gregoire, L. J., Payne, A. J., and Valdes, P. J.: Deglacial rapid sea Geophys., 42, RG1005, doi:10.1029/2003RG000128, 2004. level rises caused by ice-sheet saddle collapses, Nature, 487, Hillaire-Marcel, C., Maccali, J., Not, C., and Poirier, A.: Geochem- 219–222, doi:10.1038/nature11257, 2012. ical and isotopic tracers of Arctic sea ice sources and export with Gregoire, L. J., Otto-Bliesner, B., Valdes, P. J., and Ivanovic, R.: special attention to the Younger Dryas interval, Quaternary Sci. Abrupt Bølling warming and ice saddle collapse contributions to Rev., 79, 184–190, doi:10.1016/j.quascirev.2013.05.001, 2013. the Meltwater Pulse 1a rapid sea level rise, Geophys. Res. Lett., Hladyniuk, R. and Longstaffe, F. J.: Oxygen-isotope variations in 43, 9130–9137, doi:10.1002/2016GL070356, 2016. post-glacial Lake Ontario, Quaternary Sci. Rev., 134, 39–50, Grimaldi, S., Nardi, F., Benedetto, F. D., Istanbulluoglu, E., and doi:10.1016/j.quascirev.2016.01.002, 2016. Bras, R. L.: A physically-based method for removing pits in Hoffman, J. S., Carlson, A. E., Winsor, K., Klinkhammer, G. P., digital elevation models, Adv. Water Resour., 30, 2151–2158, LeGrande, A. N., Andrews, J. T., and Strasser, J. C.: Linking doi:10.1016/j.advwatres.2006.11.016, 2007. the 8.2 ka event and its freshwater forcing in the Labrador Sea, Gross, M. G., Karweit, M., Cronin, W. B., Schubel, J. R., Gross, Geophys. Res. Lett., 39, L18703, doi:10.1029/2012GL053047, M. G., Karweit, M., Cronin, W. B., and Schubel, J. R.: Suspended 2012. sediment discharge of the Susquehanna River to northern Chesa- Holeman, J. N.: The Sediment Yield of Major Rivers peake Bay, 1966 to 1976, Estuaries, 1, 106–110, 1978. of the World, Water Resour. Res., 4, 737–747, Guido, Z. S., Ward, D. J., and Anderson, R. S.: Pacing the post- doi:10.1029/WR004i004p00737, 1968. Last Glacial Maximum demise of the Animas Valley glacier and Hooke, R. L. and Clausen, H. B.: Wisconsin and the San Juan Mountain ice cap, Colorado, Geology, 35, 739, Holocene δ{18}O variations, Barnes Ice Cap, Canada, doi:10.1130/G23596A.1, 2007. Geol. Soc. Am. Bull., 93, 784–789, doi:10.1130/0016- Hager, B. H. and Richards, M. A.: Long-Wavelength Variations 7606(1982)93<784:WAHOVB>2.0.CO;2, 1982. in Earth’s Geoid: Physical Models and Dynamical Implications, Huang, P.-c. and Lee, K. T.: An efficient method for DEM- Philos. T. R. Soc. A, 328, 309–327, doi:10.1098/rsta.1989.0038, based overland flow routing, J. Hydrol., 489, 238–245, 1989. doi:10.1016/j.jhydrol.2013.03.014, 2013. Hansel, A. K. and Mickelson, D. M.: A reevaluation of the timing Ivanovic,´ R. F., Gregoire, L. J., Wickert, A. D., and Valdes, P. J.: and causes of high lake phases in the Lake Michigan basin, Qua- How did the North American ice Saddle Collapse impact the ternary Res., 29, 113–128, doi:10.1016/0033-5894(88)90055-5, climate 14,500 years ago?, in: AGU Fall Meeting Abstracts, 1988. PP51D–1153, American Geophysical Union, San Francisco, CA, Hanson, M. A., Lian, O. B., and Clague, J. J.: The se- USA, 15–19 December 2014. quence and timing of large late Pleistocene floods from Ivanovic, R. F., Gregoire, L. J., Kageyama, M., Roche, D. M., glacial Lake Missoula, Quaternary Sci. Rev., 31, 67–81, Valdes, P. J., Burke, A., Drummond, R., Peltier, W. R., and doi:10.1016/j.quascirev.2011.11.009, 2012. Tarasov, L.: Transient climate simulations of the deglaciation 21- Harmar, O. P. and Clifford, N. J.: Geomorphological explanation of 9 thousand years before present (version 1) – PMIP4 Core exper- the long profile of the Lower Mississippi River, Geomorphology, iment design and boundary conditions, Geosci. Model Dev., 9, 84, 222–240, doi:10.1016/j.geomorph.2006.01.045, 2007. 2563–2587, doi:10.5194/gmd-9-2563-2016, 2016. Harris, I., Jones, P., Osborn, T., and Lister, D.: Updated Ivanovic,´ R. F., Gregoire, L. J., Wickert, A. D., Valdes, P. J., and high-resolution grids of monthly climatic observations - Burke, A.: Collapse of the North American ice saddle 14,500 the CRU TS3.10 Dataset, Int. J. Climatol., 34, 623–642, years ago caused widespread cooling and reduced ocean over- doi:10.1002/joc.3711, 2014. turning circulation, Geophys. Res. Lett., submitted, 2016b. Hay, C., Mitrovica, J. X., Gomez, N., Creveling, J. R., Austermann, Jarrett, R. D. and Malde, H. E.: Paleodischarge of the late Pleis- J., and Kopp, R. E.: The sea-level fingerprints of ice-sheet col- tocene Bonneville Flood, Snake River, Idaho, computed from lapse during interglacial periods, Quaternary Sci. Rev., 87, 60– new evidence, Geol. Soc. Am. Bull., 99, 127, doi:10.1130/0016- 69, doi:10.1016/j.quascirev.2013.12.022, 2014. 7606(1987)99<127:POTLPB>2.0.CO;2, 1987. He, F.: Simulating Transiet Climate Evolution of the Last Deglacia- Jennings, A., Andrews, J., Pearce, C., Wilson, L., and Ólfasdót- tion with CCSM3, PhD thesis, University of Wisconsin – Madi- tir, S.: Detrital carbonate peaks on the Labrador shelf, a 13–7ka son, Madison, Wisconsin, USA, 2011. template for freshwater forcing from the Hudson Strait outlet of

www.earth-surf-dynam.net/4/831/2016/ Earth Surf. Dynam., 4, 831–869, 2016 864 A. D. Wickert: North American deglacial water routing

the Laurentide Ice Sheet into the subpolar gyre, Quaternary Sci. Lambeck, K., Yokoyama, Y., and Purcell, T.: Into and out of the Rev., 107, 62–80, doi:10.1016/j.quascirev.2014.10.022, 2015. Last Glacial Maximum: sea-level change during Oxygen Isotope Kammerer, J. C.: Largest Rivers in the United States, Open-File Stages 3 and 2, Quaternary Sci. Rev., 21, 343–360, 2002. Report, United States Geological Survey, Reston, Virginia, USA, Lambeck, K., Rouby, H., Purcell, A., Sun, Y., and Sambridge, M.: 1990. Sea level and global ice volumes from the Last Glacial Maximum Kaufman, D. S. and Miller, G. H.: Abrupt early Holocene (9.9– to the Holocene, P. Natl. Acad. Sci. USA, 111, 15296–15303, 9.6 ka) ice-stream advance at the mouth of Hudson Strait, doi:10.1073/pnas.1411762111, 2014. Arctic Canada, Geology, 21, 1063–1066, doi:10.1130/0091- Lavoie, C., Allard, M., and Duhamel, D.: Deglaciation landforms 7613(1993)021<1063:AEHKIS>2.3.CO;2, 1993. and C-14 chronology of the Lac Guillaume-Delisle area, eastern Kehew, A. E. and Teller, J. T.: History of late glacial runoff along the Hudson Bay: A report on field evidence, Geomorphology, 159– southwestern margin of the Laurentide Ice Sheet, Quaternary Sci. 160, 142–155, doi:10.1016/j.geomorph.2012.03.015, 2012. Rev., 13, 859–877, doi:10.1016/0277-3791(94)90006-X, 1994. Le Morzadec, K., Tarasov, L., Morlighem, M., and Seroussi, Keigwin, L. D. and Driscoll, N. W.: Deglacial floods in the Beau- H.: A new sub-grid surface mass balance and flux model for fort Sea, in: AGU Fall Meeting Abstracts, PP12B–07, American continental-scale ice sheet modelling: testing and last glacial Geophysical Union, San Francisco, CA, USA, 15–19 Decem- cycle, Geosci. Model Dev., 8, 3199–3213, doi:10.5194/gmd-8- ber 2014. 3199-2015, 2015. Keigwin, L. D., Donnelly, J. P., Cook, M. S., Driscoll, N. W., and Lehner, B., Verdin, K., and Jarvis, A.: HydroSHEDS Technical Brigham-Grette, J.: Rapid sea-level rise and Holocene climate Documentation Version 1.2, vol. 89, 2013. in the Chukchi Sea, Geology, 34, 861, doi:10.1130/G22712.1, Lemieux, J.-M., Sudicky, E. A., Peltier, W. R., and Tarasov, L.: Dy- 2006. namics of groundwater recharge and seepage over the Canadian Kendall, R. A., Mitrovica, J. X., and Milne, G. A.: On post-glacial landscape during the Wisconsinian glaciation, J. Geophys. Res., sea level–II. Numerical formulation and comparative results on 113, 1–18, doi:10.1029/2007JF000838, 2008. spherically symmetric models, Geophys. J. Int., 161, 679–706, Lemmen, D. S., Duk-Rodkin, A., and Bednarski, J. M.: Late glacial doi:10.1111/j.1365-246X.2005.02553.x, 2005. drainage systems along the northwestern margin of the Lauren- Kettner, A. J. and Syvitski, J. P. M.: HydroTrend v. 3.0: A climate- tide Ice Sheet, Quaternary Sci. Rev., 13, 805–828, 1994. driven hydrological transport model that simulates discharge and Leonard, A. G.: Pleistocene drainage changes in western North sediment load leaving a river system, Comput. & Geosci., 34, Dakota, Geol. Soc. Am. Bull., 27, 295–304, doi:10.1130/GSAB- 1170–1183, 2008. 27-295, 1916. Khider, D., Huerta, G., Jackson, C., Stott, L. D., and Emile- Leopold, L. B. and Maddock, T.: The hydraulic geometry of stream Geay, J.: A Bayesian, multivariate calibration for G lobigeri- channels and some physiographic implications, Professional Pa- noides ruber Mg/Ca, Geochem. Geophy. Geosy., 16, 2916–2932, per, United States Geological Survey, Washington, DC, 1953. doi:10.1002/2015GC005844, 2015. Lepper, K., Buell, A. W., Fisher, T. G., and Lowell, T. V.: Kim, S. J., Crowley, T. J., Erickson, D. J., Govindasamy, B., A chronology for glacial Lake Agassiz shorelines along Duffy, P. B., and Lee, B. Y.: High-resolution climate simu- Upham’s namesake transect, Quaternary Res., 80, 88–98, lation of the last glacial maximum, Clim. Dynam., 31, 1–16, doi:10.1016/j.yqres.2013.02.002, 2013. doi:10.1007/s00382-007-0332-z, 2008. Levac, E., Lewis, M., Stretch, V., Duchesne, K., and Neulieb, T.: Kirby, J. F. and Swain, C. J.: A reassessment of spectral Te esti- Evidence for meltwater drainage via the St. Lawrence River mation in continental interiors: The case of North America, J. Valley in marine cores from the Laurentian Channel at the Geophys. Res.-Sol. Ea., 114, 1–36, doi:10.1029/2009JB006356, time of the Younger Dryas, Global Planet. Change, 130, 47–65, 2009. doi:10.1016/j.gloplacha.2015.04.002, 2015. Knox, J. C.: The Mississippi River System, in: Large rivers: geo- Leventer, A., Williams, D. F., and Kennett, J. P.: Dynamics of the morphology and management, edited by: Gupta, A., 145–182, Laurentide ice sheet during the last deglaciation: evidence from John Wiley & Sons, Hoboken, NJ, USA, 2007. the Gulf of Mexico, Earth Planet. Sc. Lett., 59, 11–17, 1982. Kujau, A., Nürnberg, D., Zielhofer, C., Bahr, A., and Röhl, U.: Lewis, C. F. M. and Teller, J. T.: Glacial runoff from North Mississippi River discharge over the last ∼ 560,000 years – In- America and its possible impact on oceans and climate, dications from X-ray fluorescence core-scanning, Palaeogeogr. in: Glacier Science and Environmental Change, chap. Chap- Palaeocl., 298, 311–318, doi:10.1016/j.palaeo.2010.10.005, ter Tw, 138–150, Blackwell Publishing, Malden, MA, USA, 2010. doi:10.1002/9780470750636.ch28, 2006. Kutzbach, J. E. and Wright, H. E.: Simulation of the climate Leydet, D. J.: Eastward Routing of Glacial Lake Agassiz Runoff of 18,000 years BP: Results for the North American/North caused the Younger Dryas Cold Event, MS Thesis, Oregon State Atlantic/European sector and comparison with the geologic University, Corvallis, OR, USA, 2016. record of North America, Quaternary Sci. Rev., 4, 147–187, Licciardi, J. M., Teller, J. T., and Clark, P. U.: Freshwater rout- doi:10.1016/0277-3791(85)90024-1, 1985. ing by the Laurentide Ice Sheet during the last deglaciation, in: Laabs, B. J., Munroe, J. S., Best, L. C., and Caffee, M. W.: Timing Mechanisms of global climate change at millennial time scales, of the last glaciation and subsequent deglaciation in the Ruby edited by: Clark, P. U., Webb, R. S., and Keigwin, L. D., vol. Mountains, Great Basin, USA, Earth Planet. Sc. Lett., 361, 16– 112, Geophysical Monograph, 177–201, American Geophysical 25, doi:10.1016/j.epsl.2012.11.018, 2013. Union, doi:10.1029/gm112p0177, 1999. Liu, J., Milne, G. A., Kopp, R. E., Clark, P. U., and Shen- nan, I.: Sea-level constraints on the amplitude and source

Earth Surf. Dynam., 4, 831–869, 2016 www.earth-surf-dynam.net/4/831/2016/ A. D. Wickert: North American deglacial water routing 865

distribution of Meltwater Pulse 1A, Nat. Geosci., 9, 6–12, Meissner, K. J. and Clark, P. U.: Impact of floods versus routing doi:10.1038/ngeo2616, 2015. events on the thermohaline circulation, Geophys. Res. Lett., 33, Liu, Z., Otto-Bliesner, B. L., He, F., Brady, E. C., Tomas, R., L15704, doi:10.1029/2006GL026705, 2006. Clark, P. U., Carlson, A. E., Lynch-Stieglitz, J., Curry, W., Metz, J. M., Dowdeswell, J., and Woodworth-Lynas, C. M. T.: Sea- Brook, E., Erickson, D., Jacob, R., Kutzbach, J., and Cheng, floor scour at the mouth of Hudson Strait by deep-keeled ice- J.: Transient simulation of last deglaciation with a new mech- bergs from the Laurentide Ice Sheet, Mar. Geol., 253, 149–159, anism for Bolling-Allerod warming, Science, 325, 310–314, doi:10.1016/j.margeo.2008.05.004, 2008. doi:10.1126/science.1171041, 2009. Metz, M., Mitasova, H., and Harmon, R. S.: Efficient extraction of Livingstone, S. J., Storrar, R. D., Hillier, J. K., Stokes, C. R., Clark, drainage networks from massive, radar-based elevation models C. D., and Tarasov, L.: An ice-sheet scale comparison of eskers with least cost path search, Hydrol. Earth Syst. Sci., 15, 667– with modelled subglacial drainage routes, Geomorphology, 246, 678, doi:10.5194/hess-15-667-2011, 2011. 104–112, doi:10.1016/j.geomorph.2015.06.016, 2015. Milliman, J. D. and Farnsworth, K. L.: River discharge to the coastal Lopes, C. and Mix, A. C.: Pleistocene megafloods in the Northeast ocean: a global synthesis, Cambridge University Press, 2011. Pacific, Geology, 37, 79–82, doi:10.1130/G25025A.1, 2009. Milliman, J. D. and Meade, R. H.: World-wide delivery of river sed- Lowell, T. V., Larson, G. J., Hughes, J. D., and Denton, G. H.: iment to the oceans, J. Geol., 91, 1–21, 1983. Age verification of the Lake Gribben forest bed and the Younger Mitrovica, J. X.: Haskell [1935] revisited, J. Geophys. Res., 101, Dryas Advance of the Laurentide Ice Sheet, Can. J. Earth Sci., 555–569, doi:10.1029/95JB03208, 1996. 36, 383–393, doi:10.1139/e98-095, 1999. Mitrovica, J. X. and Forte, A. M.: A new inference of mantle viscos- Lowell, T. V., Applegate, P. J., Fisher, T. G., and Lepper, K.: What ity based upon joint inversion of convection and glacial isostatic caused the low-water phase of glacial Lake Agassiz?, Quaternary adjustment data, Earth Planet. Sc. Lett., 225, 177–189, 2004. Res., 80, 370–382, doi:10.1016/j.yqres.2013.06.002, 2013. Mitrovica, J. X. and Milne, G. A.: On post-glacial sea level: I. Gen- MacAyeal, D. R.: Binge/purge oscillations of the Laurentide Ice eral theory, Geophys. J. Int., 154, 253–267, doi:10.1046/j.1365- Sheet as a cause of the North Atlantic’s Heinrich events, Pale- 246X.2003.01942.x, 2003. oceanography, 8, 775–784, doi:10.1029/93PA02200, 1993. Monaghan, G. W.: Systematic variation in the clay-mineral compo- Maccali, J., Hillaire-Marcel, C., Carignan, J., and Reisberg, L. C.: sition of till sheets; Evidence for the Erie Interstade in the Lake Geochemical signatures of sediments documenting Arctic sea- Michigan basin, in: GSA Special Papers: Late Quaternary His- ice and water mass export through Fram Strait since the tory of the Lake Michigan Basin, edited by: Schneider, A. F. and Last Glacial Maximum, Quaternary Sci. Rev., 64, 136–151, Fraser, G. S., vol. 251, 43–50, Geological Society of America, doi:10.1016/j.quascirev.2012.10.029, 2013. Boulder, CO, USA, doi:10.1130/SPE251-p43, 1990. Mackay, J. R. and Mathews, W. H.: Geomorphology and Quaternary Monteleone, K., Dixon, E. J., and Wickert, A. D.: Lost Worlds: History of the Mackenzie River Valley near Fort Good Hope, A predictive model to locate submerged archaeological sites in N.W.T., Canada, Can. J. Earth Sci., 10, 26–41, doi:10.1139/e73- SE Alaska, USA, in: Archaeology in the Digital Era: Papers 003, 1973. from the 40th Annual Conference of Computer Applications Mandryk, C. A., Josenhans, H., Fedje, D. W., and Mathewes, and Quantitative Methods in Archaeology (CAA), Southamp- R. W.: Late Quaternary paleoenvironments of Northwestern ton, 26–29 March 2012, edited by: Earl, G., Sly, T., Chrysanthi, North America: implications for inland versus coastal migration A., Murrieta-Flores, P., Papadopoulos, C., Romanowska, I., and routes, Quaternary Sci. Rev., 20, 301–314, doi:10.1016/S0277- Wheatley, D., 1–12, Amsterdam University Press, Southampton, 3791(00)00115-3, 2001. UK, 2013. Margold, M., Stokes, C. R., Clark, C. D., and Kleman, J.: Ice Monteleone, K. R.: Lost worlds: Locating submerged archaeologi- streams in the Laurentide Ice Sheet: a new mapping inventory, cal sites in Southeast Alaska, PhD thesis, The University of New Journal of Maps, 16, 9669, doi:10.1080/17445647.2014.912036, Mexico, 2013. 2014. Montero-Serrano, J. C., Bout-Roumazeilles, V., Tribovillard, N., Margold, M., Stokes, C. R., and Clark, C. D.: Ice streams in the Sionneau, T., Riboulleau, A., Bory, A., and Flower, B.: Sedi- Laurentide Ice Sheet: Identification, characteristics and com- mentary evidence of deglacial megafloods in the northern Gulf parison to modern ice sheets, Earth-Sci. Rev., 143, 117–146, of Mexico (Pigmy Basin), Quaternary Sci. Rev., 28, 3333–3347, doi:10.1016/j.earscirev.2015.01.011, 2015. doi:10.1016/j.quascirev.2009.09.011, 2009. Marshall, S. J. and Clarke, G. K. C.: Modeling North American Moore, T. C., Walker, J. C. G., Rea, D. K., Lewis, C. F. M., Shane, freshwater runoff through the last glacial cycle, Quaternary Res., L. C. K., and Smith, A. J.: Younger Dryas Interval and out- 52, 300–315, 1999. flow from the Laurentide Ice Sheet, Paleoceanography, 15, 4–18, Matsubara, Y. and Howard, A. D.: A spatially explicit model doi:10.1029/1999PA000437, 2000. of runoff, evaporation, and lake extent: Application to mod- Mörner, N.-A. and Dreimanis, A.: The Erie interstade, Geol. Soc. ern and late Pleistocene lakes in the Great Basin region, Am. Mem., 136, 107–134, doi:10.1130/MEM136-p107, 1973. western United States, Water Resour. Res., 45, W06425, Mu, Q., Heinsch, F. A., Zhao, M., and Running, S. W.: Development doi:10.1029/2007WR005953, 2009. of a global evapotranspiration algorithm based on MODIS and McGee, D., Quade, J., Edwards, R. L., Broecker, W. S., Cheng, global meteorology data, Remote Sens. Environ., 111, 519–536, H., Reiners, P. W., and Evenson, N.: Lacustrine cave carbon- doi:10.1016/j.rse.2007.04.015, 2007. ates: Novel archives of paleohydrologic change in the Bonneville Mu, Q., Zhao, M., and Running, S. W.: Improvements to a MODIS Basin (Utah, USA), Earth Planet. Sc. Lett., 351–352, 182–194, global terrestrial evapotranspiration algorithm, Remote Sens. En- doi:10.1016/j.epsl.2012.07.019, 2012. viron., 115, 1781–1800, doi:10.1016/j.rse.2011.02.019, 2011.

www.earth-surf-dynam.net/4/831/2016/ Earth Surf. Dynam., 4, 831–869, 2016 866 A. D. Wickert: North American deglacial water routing

Munroe, J. S. and Laabs, B. J. C.: Temporal correspon- North America: Timing and controls, Geology, 31, 729–732, dence between pluvial lake highstands in the southwestern doi:10.1130/G19561.1, 2003. US and Heinrich Event 1, J. Quaternary Sci., 28, 49–58, Patterson, C. J.: Southern Laurentide ice lobes were created by ice doi:10.1002/jqs.2586, 2013. streams: Des Moines Lobe in Minnesota, USA, Sediment. Geol., Murton, J. B., Bateman, M. D., Dallimore, S. R., Teller, J. T., and 111, 249–261, doi:10.1016/S0037-0738(97)00018-3, 1997. Yang, Z.: Identification of Younger Dryas outburst flood path Patterson, C. J.: Laurentide glacial landscapes: The role of from Lake Agassiz to the Arctic Ocean, Nature, 464, 740–743, ice streams, Geology, 26, 643–646, doi:10.1130/0091- doi:10.1038/nature08954, 2010. 7613(1998)026<0643:LGLTRO>2.3.CO;2, 1998. Neteler, M., Bowman, M. H., Landa, M., and Metz, M.: GRASS Pearce, C.: The timing of the Gold Cove glacial event: A GIS: A multi-purpose open source GIS, Environ. Modell. Softw., comment on “Signature of the Gold Cove event (10.2 ka) 31, 124–130, doi:10.1016/j.envsoft.2011.11.014, 2012. in the Labrador Sea”, Quatern. Int., 377, 157–159, Nordman, M., Milne, G., and Tarasov, L.: Reappraisal of the Ånger- doi:10.1016/j.quaint.2015.02.067, 2015. man River decay time estimate and its application to determine Pearce, C., Andrews, J. T., Bouloubassi, I., Hillaire-Marcel, C., uncertainty in Earth viscosity structure, Geophys. J. Int., 201, Jennings, A. E., Olsen, J., Kuijpers, A., and Seidenkrantz, 811–822, doi:10.1093/gji/ggv051, 2015. M.-S. S.: Heinrich 0 on the east Canadian margin: Source, Nowak, K. C.: Stochastic streamflow simulation at interdecadal distribution, and timing, Paleoceanography, 30, 1613–1624, time scales and implications to water resources management in doi:10.1002/2015PA002884, 2015. the Colorado River Basin, PhD thesis, University of Colorado, Pederson, J. L., Cragun, W. S., Hidy, A. J., Rittenour, T. M., and 2011. Gosse, J. C.: Colorado River chronostratigraphy at Lee’s Ferry, Obbink, E. A., Carlson, A. E., and Klinkhammer, G. P.: Eastern Arizona, and the Colorado Plateau bull’s-eye of incision, Geol- North American freshwater discharge during the Bolling-Allerod ogy, 41, 427–430, doi:10.1130/G34051.1, 2013. warm periods, Geology, 38, 171–174, doi:10.1130/G30389.1, Pelletier, J. D.: A spatially distributed model for the long-term sus- 2010. pended sediment discharge and delivery ratio of drainage basins, Occhietti, S. and Richard, P. J.: Effet réservoir sur les âges C de la J. Geophys. Res., 117, F02028, doi:10.1029/2011JF002129, Mer de Champlain à la transition Pléistocène-Holocène: révision 2012. de la chronologie de la déglaciation au Québec méridional, Géo- Peltier, W. R., Argus, D. F., and Drummond, R.: Space geodesy graphie physique et Quaternaire, 57, 115, doi:10.7202/011308ar, constrains ice age terminal deglaciation: The global ICE- 2003. 6G\_C (VM5a)model, J. Geophys. Res.-Sol. Ea., 120, 450–487, Occhietti, S., Govare, E., Klassen, R., Parent, M., and Vincent, J.- doi:10.1002/2014JB011176, 2015. S.: Late Wisconsinan–Early Holocene deglaciation of Québec- Peltier, W. R.: Global Glacial Isostasy and the Sur- Labrador, in: Quaternary Glaciations – Extent and Chronology face of the Ice-Age Earth: The ICE-5G (VM2) Model — Part II: North America, edited by: Ehlers, J. and Gibbard, and GRACE, Annu. Rev. Earth Pl. Sc., 32, 111–149, P. L., Developments in Quaternary Sciences, 243–273, 2004. doi:10.1146/annurev.earth.32.082503.144359, 2004. Ó Cofaigh, C., Evans, D. J. A., and Smith, I. R.: Large-scale reorga- Polyak, V. J., Rasmussen, J. B., and Asmerom, Y.: Prolonged wet nization and sedimentation of terrestrial ice streams during late period in the southwestern United States through the Younger Wisconsinan Laurentide Ice Sheet deglaciation, Geol. Soc. Am. Dryas, Geology, 32, 5–8, doi:10.1130/G19957.1, 2004. Bull., 122, 743–756, doi:10.1130/B26476.1, 2010. Prairie, J. and Callejo, R.: Natural Flow And Salt Computation Orme, A. R.: Pleistocene pluvial lakes of the American West: a short Methods: Calendar Years 1971-1995, December, US Department history of research, Geological Society, London, Special Publi- of the Interior, Bureau of Reclamation, Salt Lake City, UT, and cations, 301, 51–78, doi:10.1144/SP301.4, 2008. Boulder City, NV, 2005. Oster, J. L. and Kelley, N. P.: Tracking regional and global Prince, P. S. and Spotila, J. A.: Evidence of transient topographic teleconnections recorded by western North American disequilibrium in a landward passive margin river system: knick- speleothem records, Quaternary Sci. Rev., 149, 18–33, points and paleo-landscapes of the New River basin, south- doi:10.1016/j.quascirev.2016.07.009, 2016. ern Appalachians, Earth Surf. Proc. Land., 38, 1685–1699, Oster, J. L., Ibarra, D. E., Winnick, M. J., and Maher, K.: Steering of doi:10.1002/esp.3406, 2013. westerly storms over western North America at the Last Glacial Qin, C.-z. and Zhan, L.: Parallelizing flow-accumulation calcula- Maximum, Nat. Geosci., 8, 201–205, doi:10.1038/ngeo2365, tions on graphics processing units – From iterative DEM prepro- 2015. cessing algorithm to recursive multiple-flow-direction algorithm, Overeem, I., Syvitski, J. P. M., Hutton, E. W. H., and Ket- Comput. & Geosci., 43, 7–16, doi:10.1016/j.cageo.2012.02.022, tner, A. J.: Stratigraphic variability due to uncertainty in 2012. model boundary conditions: A case-study of the New Jer- Rashid, H., Piper, D. J. W., Mansfield, C., Saint-Ange, sey Shelf over the last 40,000 years, Mar. Geol., 224, 23–41, F., and Polyak, L.: Signature of the Gold Cove event doi:10.1016/j.margeo.2005.06.044, 2005. (10.2 ka) in the Labrador Sea, Quatern. Int., 352, 212–221, Oviatt, C. G.: Chronology of Lake Bonneville, 30,000 doi:10.1016/j.quaint.2014.06.063, 2014. to 10,000 yr B.P, Quaternary Sci. Rev., 110, 166–171, Rayburn, J. A., K. Knuepfer, P. L., and Franzi, D. A.: A series of doi:10.1016/j.quascirev.2014.12.016, 2015. large, Late Wisconsinan meltwater floods through the Cham- Owen, L. A., Finkel, R. C., Minnich, R. A., and Perez, A. E.: plain and Hudson Valleys, New York State, USA, Quaternary Extreme southwestern margin of late Quaternary glaciation in Sci. Rev., 24, 2410–2419, doi:10.1016/j.quascirev.2005.02.010, 2005.

Earth Surf. Dynam., 4, 831–869, 2016 www.earth-surf-dynam.net/4/831/2016/ A. D. Wickert: North American deglacial water routing 867

Rayburn, J. A., Franzi, D. A., and Knuepfer, P. L. K.: Evi- tions, Geol. Soc. Am. Bull., 103, 1032–1048, doi:10.1130/0016- dence from the Lake Champlain Valley for a later onset of the 7606(1991)103<1032:LWPVHG>2.3.CO;2, 1991. Champlain Sea and implications for late glacial meltwater rout- Rittenour, T. M., Goble, R. J., and Blum, M. D.: An optical age ing to the North Atlantic, Palaeogeogr. Palaeocl., 246, 62–74, chronology of Late Pleistocene fluvial deposits in the northern doi:10.1016/j.palaeo.2006.10.027, 2007. lower Mississippi valley, Quaternary Sci. Rev., 22, 1105–1110, Rayburn, J. A., Cronin, T. M., Franzi, D. A., Knuepfer, P. L. K., and 2003. Willard, D. A.: Timing and duration of North American glacial Rittenour, T. M., Goble, R. J., and Blum, M. D.: Development of an lake discharges and the Younger Dryas climate reversal, Quater- OSL chronology for Late Pleistocene channel belts in the lower nary Res., 75, 541–551, doi:10.1016/j.yqres.2011.02.004, 2011. Mississippi valley, USA, Quaternary Sci. Rev., 24, 2539–2554, Rech, J. A., Nekola, J. C., and Pigati, J. S.: Radiocarbon ages of doi:10.1016/j.quascirev.2005.03.011, 2005. terrestrial gastropods extend duration of ice-free conditions at the Rittenour, T. M., Blum, M. D., and Goble, R. J.: Fluvial evolution of Two Creeks forest bed, Wisconsin, USA, Quaternary Res., 77, the lower Mississippi River valley during the last 100 k.y. glacial 289–292, doi:10.1016/j.yqres.2011.11.007, 2012. cycle: Response to glaciation and sea-level change, Geol. Soc. Refsnider, K. A., Laabs, B. J. C., Plummer, M. A., Mickelson, Am. Bull., 119, 586–608, doi:10.1130/B25934.1, 2007. D. M., Singer, B. S., and Caffee, M. W.: Last glacial maximum Roberts, G. G., White, N. J., Martin-Brandis, G. L., and Crosby, climate inferences from cosmogenic dating and glacier modeling A. G.: An uplift history of the Colorado Plateau and its surround- of the western Uinta ice field, Uinta Mountains, Utah, Quaternary ings from inverse modeling of longitudinal river profiles, Tecton- Res., 69, 130–144, doi:10.1016/j.yqres.2007.10.014, 2008. ics, 31, 1–2, doi:10.1029/2012TC003107, 2012. Refsnider, K. A., Brugger, K. A., Leonard, E. M., Mccalpin, Rohling, E. J., Grant, K., Hemleben, C. H., Siddall, M., Hoogakker, J. P., and Armstrong, P. P.: Last Glacial Maximum equilibrium- B. A. A., Bolshaw, M., and Kucera, M.: High rates of sea-level line altitude trends and precipitation patterns in the Sangre de rise during the last interglacial period, Nat. Geosci., 1, 38–42, Cristo Mountains, southern Colorado, USA, Boreas, 38, 663– doi:10.1038/ngeo.2007.28, 2007. 678, doi:10.1111/j.1502-3885.2009.00097.x, 2009. Ross, M., Campbell, J. E., Parent, M., and Adams, R. S.: Reimer, P. J., Bard, E., Bayliss, A., Beck, J. W., Blackwell, P. G., Palaeo-ice streams and the subglacial landscape mosaic of the Ramsey, C. B., Grootes, P. M., Guilderson, T. P., Haflidason, H., North American mid-continental prairies, Boreas, 38, 421–439, Hajdas, I., Hatté, C., Heaton, T. J., Hoffmann, D. L., Hogg, A. G., doi:10.1111/j.1502-3885.2009.00082.x, 2009. Hughen, K. A., Kaiser, K. F., Kromer, B., Manning, S. W., Niu, Roy, M., Dell’Oste, F., Veillette, J. J., de Vernal, A., Hélie, M., Reimer, R. W., Richards, D. A., Scott, E. M., Southon, J. R., J.-F., and Parent, M.: Insights on the events surrounding Staff, R. A., Turney, C. S. M., van der Plicht, J., and van der the final drainage of Lake Ojibway based on James Bay Plicht, J.: IntCal13 and Marine13 Radiocarbon Age Calibration stratigraphic sequences, Quaternary Sci. Rev., 30, 682–692, Curves 0–50,000 Years cal BP, Radiocarbon, 55, 1869–1887, doi:10.1016/j.quascirev.2010.12.008, 2011. doi:10.2458/azu_js_rc.55.16947, 2013. Rutt, I. C., Hagdorn, M., Hulton, N. R. J., and Payne, A. J.: The Remenda, V. H., Cherry, J. A., and Edwards, T. W. D.: Isotopic Glimmer community ice sheet model, J. Geophys. Res., 114, Composition of Old Ground Water from Lake Agassiz: Impli- F02004, doi:10.1029/2008JF001015, 2009. cations for Late Pleistocene Climate, Science, 266, 1975–1978, Schmidt, J. C., Schmidt, B. J. C., and Schmidt, J. C.: A Watershed doi:10.1126/science.266.5193.1975, 1994. Perspective of Changes in Streamflow , Sediment Supply , and Reneau, S. L. and Dethier, D. P.: Late Pleistocene landslide- Geomorphology of the Colorado River, in: Proceedings of the dammed lakes along the Rio Grande, White Rock Canyon, Colorado River Basin Science and Resource Management Sym- New Mexico, Geol. Soc. Am. Bull., 108, 1492–1507, posium, 18–20 November 2008, Scottsdale, Arizona, edited by: doi:10.1130/0016-7606(1996)108<1492:LPLDLA>2.3.CO;2, Melis, T. S., Hamill, J. F., Bennett, G. E., Lewis G. Coggins, J., 1996. Grams, P. E., Kennedy, T. A., Kubly, D. M., and Ralston, B. E., Reusser, L. J., Bierman, P. R., Pavich, M. J., Zen, E.-a., Larsen, vol. 2009, USGS Scientific Investigations Report, 51–76, United J., and Finkel, R.: Rapid late Pleistocene incision of At- States Geological Survey, Scottsdale, Arizona, USA, 2010. lantic passive-margin river gorges, Science, 305, 499–502, Schneider-Vieira, F., Baker, R., and Lawrence, M.: The Estuaries doi:10.1126/science.1097780, 2004. of Hudson Bay: A Case Study of the Physical and Biological Reusser, L., Bierman, P., and Pavich, M.: An episode of rapid Characteristics of Selected Sites, North/South Consultants Inc., bedrock channel incision during the last glacial cycle, measured Winnipeg, Manitoba, Canada, 1994. with 10Be, Am. J. Sci., 306, 69–102, 2006. Schwanghart, W. and Scherler, D.: Short Communication: Topo- Richard, P. J. H. and Occhietti, S.: 14C chronology for Toolbox 2 – MATLAB-based software for topographic analysis ice retreat and inception of Champlain Sea in the St. and modeling in Earth surface sciences, Earth Surf. Dynam., 2, Lawrence Lowlands, Canada, Quaternary Res., 63, 353–358, 1–7, doi:10.5194/esurf-2-1-2014, 2014. doi:10.1016/j.yqres.2005.02.003, 2005. Sionneau, T., Bout-Roumazeilles, V., Biscaye, P. E., Van Vliet- Ridge, J. C.: Shed Brook Discontinuity and Little Falls Lanoe, B., and Bory, A.: Clay mineral distributions in and around Gravel: Evidence for the Erie interstade in central New the Mississippi River watershed and Northern Gulf of Mexico: York, Geol. Soc. Am. Bull., 109, 652–665, doi:10.1130/0016- sources and transport patterns, Quaternary Sci. Rev., 27, 1740– 7606(1997)109<0652:SBDALF>2.3.CO;2, 1997. 1751, doi:10.1016/j.quascirev.2008.07.001, 2008. Ridge, J. C., Franzi, D. A., and Muller, E. H.: Late Wis- Sionneau, T., Bout-Roumazeilles, V., Flower, B. P., Bory, A., Tri- consinan, pre-Valley Heads glaciation in the western Mo- bovillard, N., Kissel, C., Van Vliet-Lanoë, B., and Montero Ser- hawk Valley, central New York, and its regional implica- rano, J. C.: Provenance of freshwater pulses in the Gulf of Mex-

www.earth-surf-dynam.net/4/831/2016/ Earth Surf. Dynam., 4, 831–869, 2016 868 A. D. Wickert: North American deglacial water routing

ico during the last deglaciation, Quaternary Res., 74, 235–245, Arctic trigger for the Younger Dryas, Quaternary Sci. Rev., 25, doi:10.1016/j.yqres.2010.07.002, 2010. 659–688, doi:10.1016/j.quascirev.2005.12.006, 2006. Smith, D. G.: Glacial lake McConnell: Paleogeography, age, dura- Tarasov, L., Dyke, A. S., Neal, R. M., and Peltier, W. R.: tion, and associated river deltas, mackenzie river basin, western A data-calibrated distribution of deglacial chronologies Canada, Quaternary Sci. Rev., 13, 829–843, doi:10.1016/0277- for the North American ice complex from glaciologi- 3791(94)90004-3, 1994. cal modeling, Earth Planet. Sc. Lett., 315–316, 30–40, Smith, R. S., Gregory, J. M., and Osprey, A.: A description of doi:10.1016/j.epsl.2011.09.010, 2012. the FAMOUS (version XDBUA) climate model and control Taylor, M., Hendy, I., and Pak, D.: Deglacial ocean warming run, Geosci. Model Dev., 1, 53–68, doi:10.5194/gmd-1-53-2008, and marine margin retreat of the Cordilleran Ice Sheet in 2008. the North Pacific Ocean, Earth Planet. Sc. Lett., 403, 89–98, Spero, H. J., Eggins, S. M., Russell, A. D., Vetter, L., Kilburn, doi:10.1016/j.epsl.2014.06.026, 2014. M. R., and Hönisch, B.: Timing and mechanism for intratest Teller, J. T.: Preglacial (Teays) and Early Glacial Drainage Mg/Ca variability in a living planktic foraminifer, Earth Planet. in the Cincinnati Area, Ohio, Kentucky, and Indiana, Sc. Lett., 409, 32–42, doi:10.1016/j.epsl.2014.10.030, 2015. Geol. Soc. Am. Bull., 84, 3677, doi:10.1130/0016- Stanford, S. D.: Onshore record of Hudson River drainage to the 7606(1973)84<3677:PTAEGD>2.0.CO;2, 1973. continental shelf from the late Miocene through the late Wis- Teller, J. T.: Meltwater and precipitation runoff to the North At- consinan deglaciation, USA: synthesis and revision, Boreas, 39, lantic, Arctic, and Gulf of Mexico from the Laurentide Ice Sheet 1–17, doi:10.1111/j.1502-3885.2009.00106.x, 2010. and adjacent regions during the Younger Dryas, Paleoceanogra- Stanford, S. D., Witte, R. W., Braun, D. D., and Ridge, J. C.: Qua- phy, 5, 897–905, doi:10.1029/PA005i006p00897, 1990a. ternary fluvial history of the Delaware River, New Jersey and Teller, J. T.: Volume and routing of late-glacial runoff from the Pennsylvania, USA: The effects of glaciation, glacioisostasy, and southern Laurentide Ice Sheet, Quaternary Res., 34, 12–23, eustasy on a proglacial river system, Geomorphology, 264, 12– doi:10.1016/0033-5894(90)90069-W, 1990b. 28, doi:10.1016/j.geomorph.2016.04.002, 2016. Teller, J. T.: Lake Agassiz during the Younger Dryas, Quaternary Stokes, C. R., Clark, C. D., and Storrar, R.: Major changes in ice Res., 80, 361–369, doi:10.1016/j.yqres.2013.06.011, 2013. stream dynamics during deglaciation of the north-western margin Teller, J. T. and Leverington, D. W.: Glacial Lake Agassiz: of the Laurentide Ice Sheet, Quaternary Sci. Rev., 28, 721–738, A 5000 yr history of change and its relationship to the doi:10.1016/j.quascirev.2008.07.019, 2009. δ18O record of Greenland, Geol. Soc. Am. Bull., 116, 729, Stokes, C. R., Tarasov, L., Blomdin, R., Cronin, T. M., Fisher, doi:10.1130/B25316.1, 2004. T. G., Gyllencreutz, R., Hättestrand, C., Heyman, J., Hindmarsh, Teller, J. T., Leverington, D. W., and Mann, J. D.: Freshwater out- R. C., Hughes, A. L., Jakobsson, M., Kirchner, N., Livingstone, bursts to the oceans from glacial Lake Agassiz and their role in S. J., Margold, M., Murton, J. B., Noormets, R., Peltier, W. R., climate change during the last deglaciation, Quaternary Sci. Rev., Peteet, D. M., Piper, D. J., Preusser, F., Renssen, H., Roberts, 21, 879–887, doi:10.1016/S0277-3791(01)00145-7, 2002. D. H., Roche, D. M., Saint-Ange, F., Stroeven, A. P., and Teller, Tesauro, M., Kaban, M. K., and Cloetingh, S. A. P. L.: Global J. T.: On the reconstruction of palaeo-ice sheets: Recent ad- strength and elastic thickness of the lithosphere, Global Planet. vances and future challenges, Quaternary Sci. Rev., 125, 15–49, Change, 90–91, 51–57, doi:10.1016/j.gloplacha.2011.12.003, doi:10.1016/j.quascirev.2015.07.016, 2015. 2012. Storrar, R. D., Stokes, C. R., and Evans, D. J.: A map of large Cana- Tight, W. G.: Drainage modifications in southeastern Ohio and ad- dian eskers from Landsat satellite imagery, Journal of Maps, 9, jacent parts of West Virginia and Kentucky, US Geological Sur- 456–473, doi:10.1080/17445647.2013.815591, 2013. vey Professional Paper, US Geological Survey, Reston, Virginia, Stroup, J. S., Lowell, T. V., and Breckenridge, A.: A model for USA, 1903. the demise of large, glacial Lake Ojibway, Ontario and Quebec, Todd, J. E.: The Pleistocene history of the Missouri River, Science, J. Paleolimnol., 50, 105–121, doi:10.1007/s10933-013-9707-9, 39, 263, 1914. 2013. Tripsanas, E. K., Bryant, W. R., Slowey, N. C., Bouma, A. H., Svensson, A., Andersen, K. K., Bigler, M., Clausen, H. B., Dahl- Karageorgis, A. P., and Berti, D.: Sedimentological history of Jensen, D., Davies, S. M., Johnsen, S. J., Muscheler, R., Parrenin, Bryant Canyon area, northwest Gulf of Mexico, during the last F., Rasmussen, S. O., Röthlisberger, R., Seierstad, I., Steffensen, 135 kyr (Marine Isotope Stages 1–6): A proxy record of Mis- J. P., and Vinther, B. M.: A 60 000 year Greenland stratigraphic sissippi River discharge, Palaeogeogr. Palaeocl., 246, 137–161, ice core chronology, Clim. Past, 4, 47–57, doi:10.5194/cp-4-47- doi:10.1016/j.palaeo.2006.10.032, 2007. 2008, 2008. Tripsanas, E. K., Karageorgis, A. P., Panagiotopoulos, I. P., Syvitski, J. P. M., Peckham, S. D., Hilberman, R., and Mul- Koutsopoulou, E., Kanellopoulos, T. D., Bryant, W. R., and der, T.: Predicting the terrestrial flux of sediment to the global Slowey, N. C.: Paleoenvironmental and paleoclimatic implica- ocean: a planetary perspective, Sediment. Geol., 162, 5–24, tions of enhanced Holocene discharge from the Mississippi River doi:10.1016/S0037-0738(03)00232-X, 2003. Based on the sedimentology and geochemistry of a deep core Tarasov, L. and Peltier, W. R.: Arctic freshwater forcing of (JPC-26) from the Gulf Of Mexico, PALAIOS, 28, 623–636, the Younger Dryas cold reversal, Nature, 435, 662–665, doi:10.2110/palo.2012.p12-060r, 2014. doi:10.1038/nature03617, 2005. Tucker, G. E. and Hancock, G. R.: Modelling landscape evolution, Tarasov, L. and Peltier, W. R. R.: A calibrated deglacial drainage Earth Surf. Proc. Land., 35, 28–50, doi:10.1002/esp.1952, 2010. chronology for the North American continent: evidence of an Tucker, G. E. and Whipple, K. X.: Topographic outcomes predicted by stream erosion models: Sensitivity analysis

Earth Surf. Dynam., 4, 831–869, 2016 www.earth-surf-dynam.net/4/831/2016/ A. D. Wickert: North American deglacial water routing 869

and intermodel comparison, J. Geophys. Res., 107, 1–16, Waitt, R. B. J.: About forty last-glacial Lake Missoula jökulhlaups doi:10.1029/2001JB000162, 2002. through southern Washington, J. Geol., 88, 653–679, 1980. Tushingham, A. M. and Peltier, W. R.: Ice-3G: A new global model Wall, G. R., Nystrom, E. A., and Litten, S.: Suspended Sedi- of Late Pleistocene deglaciation based upon geophysical predic- ment Transport in the Freshwater Reach of the Hudson River tions of post-glacial relative sea level change, J. Geophys. Res., Estuary in Eastern New York, Estuar. Coast., 31, 542–553, 96, 4497, doi:10.1029/90JB01583, 1991. doi:10.1007/s12237-008-9050-y, 2008. Ullman, D. J., LeGrande, A. N., Carlson, A. E., Anslow, F. S., Wickert, A. D.: Impacts of Pleistocene glaciation and its geophysi- and Licciardi, J. M.: Assessing the impact of Laurentide Ice cal effects on North American river systems, PhD thesis, Univer- Sheet topography on glacial climate, Clim. Past, 10, 487–507, sity of Colorado Boulder, 2014. doi:10.5194/cp-10-487-2014, 2014. Wickert, A. D.: Reconstruction of North American Drainage Basins Ullman, D. J., Carlson, A. E., Anslow, F. S., Legrande, A. N., and and River Discharge Since the Last Glacial Maximum, Data Re- Licciardi, J. M.: Laurentide ice-sheet instability during the last pos. U M, doi:10.13020/D6D01H, 2016. deglaciation, Nat. Geosci., 8, 534–537, doi:10.1038/NGEO2463, Wickert, A. D., Mitrovica, J. X., Williams, C., and Ander- 2015. son, R. S.: Gradual demise of a thin southern Laurentide ice Upham, W.: Lake Agassiz: a chapter in glacial geology, Bulletin of sheet recorded by Mississippi drainage, Nature, 502, 668–671, the Minnesota Academy of Natural Sciences, 2, 290–314, 1883. doi:10.1038/nature12609, 2013. U.S. Bureau of Reclamation: Report on the Water Supply of the Willgoose, G.: Mathematical Modeling of Whole Land- Lower Colorado River, Project Planning Report, 1952. scape Evolution, Annu. Rev. Earth Pl. Sc., 33, 443–459, van Andel, T. H.: Recent marine sediments of the Gulf of California, doi:10.1146/annurev.earth.33.092203.122610, 2005. in: Mar. Geol. of the Gulf of California: A Symposium, edited by: Williams, C., Flower, B. P., and Hastings, D. W.: Deglacial van Andel, T. J. and Shor, G. G., Memoir, 216–310, American abrupt climate change in the Atlantic Warm Pool: A Association of Petroleum Geologists, 1964. Gulf of Mexico perspective, Paleoceanography, 115, PA4221, van Hise, C. R. and Leith, C. K.: The geology of the Lake Supe- doi:10.1029/2010PA001928, 2010. rior region, U.S. Government Printing Ofﬁce, Washington, DC, Williams, C., Flower, B. P., and Hastings, D. W.: Seasonal Lau- available at: http://collections.mnhs.org/MNHistoryMagazine/ rentide Ice Sheet melting during the “Mystery Interval” (17.5– articles/13/v13i04p403-407.pdf (last access: 1 February 2014), 14.5 ka), Geology, 40, 955–958, doi:10.1130/G33279.1, 2012. 1911. Wright, H. E.: Tunnel valleys, glacial surges, and subglacial hydrol- Vázquez Riveiros, N., Govin, A., Waelbroeck, C., Mackensen, A., ogy of the Superior Lobe, Minnesota, Geol. Soc. Am. Mem., 136, Michel, E., Moreira, S., Bouinot, T., Caillon, N., Orgun, A., and 251–276, 1973. Brandon, M.: Mg/Ca thermometry in planktic foraminifera: Im- Wright, H. E.: Synthesis: the land south of the ice sheets, in: North proving paleotemperature estimations for G. bulloides and N. America and Adjacent Oceans During the Last Deglaciation, pachyderma left, Geochem. Geophy. Geosy., 17, 1249–1264, edited by: Ruddiman, W. F. and Wright, H. E. J., Geology of doi:10.1002/2015GC006234, 2016. North America, 479–488, Geological Society of America, Boul- Veillette, J. J.: Evolution and paleohydrology of glacial Lakes der, Colorado, USA, 1987. Barlow and Ojibway, Quaternary Sci. Rev., 13, 945–971, Xie, P. and Arkin, P. A.: Global Precipitation: A 17- doi:10.1016/0277-3791(94)90010-8, 1994. Year Monthly Analysis Based on Gauge Observations, Vening Meinesz, F. A.: Une nouvelle methode pour la reduc- Satellite Estimates, and Numerical Model Outputs, B. tion isostatique regionale de l’intensite de la pesanteur, Bulletin Am. Meteorol. Soc., 78, 2539–2558, doi:10.1175/1520- Géodésique (1922–1941), 29, 33–51, 1931. 0477(1997)078<2539:GPAYMA>2.0.CO;2, 1997. Ver Steeg, K.: The Teays River, Ohio J. Sci., 46, 297–307, 1946. Yeager, S. G., Shields, C. A., Large, W. G., and Hack, Vetter, L.: Combined Stable Isotope and Trace Element Analyses J. J.: The low-resolution CCSM3, J. Climate, 19, 2545–2566, on Single Planktic Foraminifer Shells: Insights from Live Cul- doi:10.1175/JCLI3744.1, 2006. ture Experiments and Paleoceanographic Applications, PhD the- Zuffa, G., Normark, W., Serra, F., and Brunner, C.: Turbidite sis, University of California, Davis, 2013. megabeds in an oceanic rift valley recording jökulhlaups of late Vincent, J.-S. and Hardy, L.: L’évolution et l’extension des lacs Pleistocene glacial lakes of the western United States, J. Geol., glaciaires Barlow et Ojibway en territoire québécois, Géographie 108, 253–274, 2000. physique et Quaternaire, 31, 357–372, doi:10.7202/1000283ar, 1977. Wagner, J. D. M., Cole, J. E., Beck, J. W., Patchett, P. J., Henderson, G. M., and Barnett, H. R.: Moisture variability in the southwestern United States linked to abrupt glacial climate change, Nat. Geosci., 3, 110–113, doi:10.1038/ngeo707, 2010.

www.earth-surf-dynam.net/4/831/2016/ Earth Surf. Dynam., 4, 831–869, 2016